{"state":{"revisions":[{"id":"3557","callsign":"HG","title":"extensions: new closehead module for closing arbitrary heads","author":"joerg.sonnenberger","summary":"``hg close-head`` allows closing arbitrary heads. It is equivalent to\nchecking out the given revisions and commit an empty change with\n``hg commit --close-branch``.","testPlan":"","lineCount":"145","dependsOn":[],"reviewers":[],"ccs":["indygreg","pulkit","mercurial-devel"],"actions":[{"type":"comment","comment":"Adjusted for most of the review comments. Added a test that books are left alone, since they will not be active in the interesting cases. I'm not touching any existing phases, so I'm not sure if I should do anything about phase testing?","author":"joerg.sonnenberger","id":"71216","dateCreated":"1537525563","dateModified":"1537525563"},{"type":"update","diffId":"11254","author":"joerg.sonnenberger","id":"71213","dateCreated":"1537525502","dateModified":"1537525502"},{"type":"update","diffId":"11253","author":"joerg.sonnenberger","id":"71212","dateCreated":"1537525307","dateModified":"1537525307"},{"type":"inline","comment":"revs can be passed as opts['rev'] too now.","replyTo":null,"isNewFile":"1","line":"55","lineLength":"0","path":"hgext\/closehead.py","diffId":"9400","author":"pulkit","id":"61802","dateCreated":"1532371616","dateModified":"1532371616"},{"type":"comment","comment":"Sorry for getting late to review this. Found a minor nit and also looks like you missed adding tests for bookmark movement and phase preservation.","author":"pulkit","id":"61801","dateCreated":"1532371616","dateModified":"1532371616"},{"type":"update","diffId":"9400","author":"joerg.sonnenberger","id":"60485","dateCreated":"1530483867","dateModified":"1530483867"},{"type":"update","diffId":"9399","author":"joerg.sonnenberger","id":"60484","dateCreated":"1530483326","dateModified":"1530483326"},{"type":"inline","comment":"I don't think this will ever go the else part.","replyTo":null,"isNewFile":"1","line":"79","lineLength":"0","path":"hgext\/closehead.py","diffId":"8963","author":"pulkit","id":"58291","dateCreated":"1528742729","dateModified":"1528742729"},{"type":"inline","comment":"please add an `opts = pycompat.byteskwargs(opts)` call here for python 3 support.","replyTo":null,"isNewFile":"1","line":"53","lineLength":"0","path":"hgext\/closehead.py","diffId":"8963","author":"pulkit","id":"58290","dateCreated":"1528742729","dateModified":"1528742729"},{"type":"inline","comment":"close is too generic, I don't feel confident enough to have this as a command name. Can we drop that for now if you don't feel to strong?","replyTo":null,"isNewFile":"1","line":"33","lineLength":"0","path":"hgext\/closehead.py","diffId":"8963","author":"pulkit","id":"58289","dateCreated":"1528742729","dateModified":"1528742729"},{"type":"inline","comment":"same, don't create a new inside existing repo, cd back and then create one.","replyTo":null,"isNewFile":"1","line":"43","lineLength":"0","path":"tests\/test-close-head.t","diffId":"8963","author":"pulkit","id":"58288","dateCreated":"1528742729","dateModified":"1528742729"},{"type":"inline","comment":"init a new repo and cd into that. don't work in top level $TESTTMP directory.","replyTo":null,"isNewFile":"1","line":"1","lineLength":"0","path":"tests\/test-close-head.t","diffId":"8963","author":"pulkit","id":"58287","dateCreated":"1528742729","dateModified":"1528742729"},{"type":"inline","comment":"we don't start with uppercase in error messages","replyTo":null,"isNewFile":"1","line":"69","lineLength":"0","path":"hgext\/closehead.py","diffId":"8963","author":"pulkit","id":"58286","dateCreated":"1528742729","dateModified":"1528742729"},{"type":"inline","comment":"nit: how about using context manager?","replyTo":null,"isNewFile":"1","line":"73","lineLength":"0","path":"hgext\/closehead.py","diffId":"8963","author":"pulkit","id":"58285","dateCreated":"1528742729","dateModified":"1528742729"},{"type":"comment","comment":"Thanks for taking this out into a separate extension. Code looks mostly good to me. It feels to me that we need some more tests:\n\n  * passing multiple revisions when some are head and some not\n  * closing a head which has secret phase\n  * checking for bookmark movement\n\nI know the behavior in couple of the above points but having tests will help us in future when we change some behavior, like showing warning for revs which are not heads and closing others.\n\nAlso, wherever we pass revisions, we have support for `--rev` flag too. Can you add that too?","author":"pulkit","id":"58284","dateCreated":"1528742729","dateModified":"1528742729"},{"type":"update","diffId":"8963","author":"joerg.sonnenberger","id":"57896","dateCreated":"1528041430","dateModified":"1528041430"},{"type":"comment","comment":"I'm -0.25 on this.\n\nBut I don't think we should add an argument to `hg heads` or `hg branch` because I don't like a) overloading commands to have multiple meanings b) making commands read-write and read-only. I think commands should do one thing and do them well. There should also be a set of \/\/safe\/\/ commands that can be executed without meaningful side-effects.\n\nFollowing that logic, a dedicated command is justified.\n\nBut I just don't feel like closing heads that aren't checked out is a frequent enough operation to justify a dedicated command.\n\nAs a potential compromise, how about `hg commit --close-branch-rev <REV>` that does the branch closing without needing a working directory? But this seemingly violates the expectation of `hg commit`, which is that it operates on the working directory. Gah - now I think I talked myself into a dedicated command.\n\nGood UI design is hard.","author":"indygreg","id":"57734","dateCreated":"1527728586","dateModified":"1527728586"},{"type":"comment","comment":"If we want to go with a `--close` flag, I think the most natural place would actually be `hg heads`. This is not really a branch operation after all, but about cutting off heads. At the same time, none of those options would allow specifying a commit message naturally.","author":"joerg.sonnenberger","id":"57441","dateCreated":"1527248702","dateModified":"1527248702"},{"type":"comment","comment":">>! In D3557#57279, @joerg.sonnenberger wrote:\n> It doesn't seem to fit the argument schema for `hg branch` at all. I'm open for better places, but I couldn't think of a consistent place.\n\nThe branch command has a `--rev` flag which we can use. Something like `hg branch --rev <revs> --close`. What do you think?","author":"pulkit","id":"57440","dateCreated":"1527247871","dateModified":"1527247871"},{"type":"inline","comment":"we use all lowercase error messages, so this should be \"no commit ..\".","replyTo":null,"isNewFile":"1","line":"1508","lineLength":"0","path":"mercurial\/commands.py","diffId":"8680","author":"pulkit","id":"57439","dateCreated":"1527247802","dateModified":"1527247802"},{"type":"comment","comment":"The code looks mostly good to me. Needs tests for following cases:\n\n  * when working directory is dirty\n  * when we are in an uncommitted merge state\n  * there is some bookmark movement code, tests for that","author":"pulkit","id":"57438","dateCreated":"1527247802","dateModified":"1527247802"},{"type":"comment","comment":"It doesn't seem to fit the argument schema for `hg branch` at all. I'm open for better places, but I couldn't think of a consistent place.","author":"joerg.sonnenberger","id":"57279","dateCreated":"1526936584","dateModified":"1526936584"},{"type":"comment","comment":"> `hg close-branch` allows closing arbitrary heads. It is equivalent to\n> checking out the given revisions and commit an empty change with\n> `hg commit --close-branch`.\n\nI am -1 on having a new command for this. How about adding this as a `--close` flag to `hg branch` command?","author":"pulkit","id":"57182","dateCreated":"1526845960","dateModified":"1526845960"},{"type":"update","diffId":"8680","author":"joerg.sonnenberger","id":"56233","dateCreated":"1526254935","dateModified":"1526254935"}],"dateCreated":"1526254935","dateModified":"1537525563","status":"Needs Review"},{"id":"4312","callsign":"HG","title":"New bookflow extension for bookmark-based branching","author":"idlsoft","summary":"This extension should be helpful for feature branches - based workflows.\nAt my company we first considered branches, but weren't sure about creating a lot of permanent objects.\nWe tried bookmarks, but found some scenarios to be difficult to control.\n\nThe main problem, was the active bookmark being moved on update.\nDisabling that made everything a lot more predictable.\n\nBookmarks move on commit, and updating means switching between them.\n\nThe extension also implements a few minor features to better guide the workflow:\n  - hg bookmark NAME can be ambiguous (create or move), unlike hg branch. The extension requires -rev to move.\n  - require an active bookmark on commit.\n  - some bookmarks can be protected (like @), useful for teams, that require code reviews.\n  - block creation of new branches.\n\nThe initial implementation requires no changes in the core, but it does rely on some implementation details.\nI thought it may be useful to discuss the functionality first, and then focus making the code more robust. ","testPlan":"","lineCount":"375","dependsOn":[],"reviewers":["pulkit"],"ccs":["jwatt","pulkit","mercurial-devel"],"actions":[{"type":"comment","comment":"The code looks mostly good. Since this is something which changes existing bookmark behavior, I will let one more pair of eyes to look before getting pushed.","author":"pulkit","id":"71189","dateCreated":"1537479138","dateModified":"1537479138"},{"type":"accept","author":"pulkit","id":"71188","dateCreated":"1537479138","dateModified":"1537479138"},{"type":"update","diffId":"10789","author":"idlsoft","id":"68532","dateCreated":"1536156265","dateModified":"1536156265"},{"type":"update","diffId":"10754","author":"idlsoft","id":"68319","dateCreated":"1536082465","dateModified":"1536082465"},{"type":"inline","comment":"Also, this way it looks more like a cheatsheet","replyTo":"68080","isNewFile":"1","line":"14","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10508","author":"idlsoft","id":"68280","dateCreated":"1536075385","dateModified":"1536075385"},{"type":"inline","comment":"I uppercased it to distinguish the parameter from the command. \nI thought hg uses this convention:\n```\nhg bookmarks [OPTIONS]... [NAME]...\n```","replyTo":"68080","isNewFile":"1","line":"14","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10508","author":"idlsoft","id":"68111","dateCreated":"1535864427","dateModified":"1535864427"},{"type":"inline","comment":"oh, that was done to see where the change was made, you'd have lines like:\n```\ntest a\ntest a\ntest b\n...\n```","replyTo":"68088","isNewFile":"1","line":"2","lineLength":"0","path":"tests\/test-bookflow.t","diffId":"10508","author":"idlsoft","id":"68110","dateCreated":"1535864427","dateModified":"1535864427"},{"type":"inline","comment":"Yes, I do this in a few other places, when I don't care about the output, only the exit code.\nIt helps shorten the testcase to just the relevant pieces.","replyTo":"68087","isNewFile":"1","line":"29","lineLength":"0","path":"tests\/test-bookflow.t","diffId":"10508","author":"idlsoft","id":"68109","dateCreated":"1535864427","dateModified":"1535864427"},{"type":"inline","comment":"You're probably right :)","replyTo":"68089","isNewFile":"1","line":"3","lineLength":"0","path":"tests\/test-bookflow.t","diffId":"10508","author":"idlsoft","id":"68108","dateCreated":"1535864427","dateModified":"1535864427"},{"type":"inline","comment":"They can, but I left them to make sure people aren't tempted to add a hook.","replyTo":"68086","isNewFile":"1","line":"89","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10508","author":"idlsoft","id":"68107","dateCreated":"1535864427","dateModified":"1535864427"},{"type":"inline","comment":"Did you mean `hg help bookflow`?","replyTo":"68083","isNewFile":"1","line":"82","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10508","author":"idlsoft","id":"68106","dateCreated":"1535864427","dateModified":"1535864427"},{"type":"inline","comment":"For me personally \"out of sync\" is more obvious, it underlines the fact that you won't be able to do commits.","replyTo":"68085","isNewFile":"1","line":"72","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10508","author":"idlsoft","id":"68105","dateCreated":"1535864427","dateModified":"1535864427"},{"type":"comment","comment":"I added some inline comments.\nA test for `hg book EXISTING` is indeed missing, there is, however, one for not moving the bookmark on update, it's line 52:  ` (leaving bookmark X)`.\nHopefully, I'll be able to make some changes later in the week.\nWas not aware about the wiki page, this indeed seems to be addressing the same problem. At least that's how it started., and later evolved into something that helps you follow the workflow.","author":"idlsoft","id":"68104","dateCreated":"1535864427","dateModified":"1535864427"},{"type":"inline","comment":"nit: -R flag can be used.","replyTo":null,"isNewFile":"1","line":"255","lineLength":"0","path":"tests\/test-bookflow.t","diffId":"10508","author":"pulkit","id":"68095","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"Can we just show the whole output. We are completely fine with more redundant lines.","replyTo":null,"isNewFile":"1","line":"258","lineLength":"0","path":"tests\/test-bookflow.t","diffId":"10508","author":"pulkit","id":"68094","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"Same, let's not push output to \/dev\/null.","replyTo":null,"isNewFile":"1","line":"275","lineLength":"0","path":"tests\/test-bookflow.t","diffId":"10508","author":"pulkit","id":"68093","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"Same, let's not push output to \/dev\/null.","replyTo":null,"isNewFile":"1","line":"271","lineLength":"0","path":"tests\/test-bookflow.t","diffId":"10508","author":"pulkit","id":"68092","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"Opinion: these assert_clean and make_changes are not clean ways. Maybe the global -R flag can be used to commit without changing directories.","replyTo":null,"isNewFile":"1","line":"123","lineLength":"0","path":"tests\/test-bookflow.t","diffId":"10508","author":"pulkit","id":"68091","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"I suggest printing the complete output of `hg pull` and remove the grep.","replyTo":null,"isNewFile":"1","line":"110","lineLength":"0","path":"tests\/test-bookflow.t","diffId":"10508","author":"pulkit","id":"68090","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"sorry, but is it like `hg status`?","replyTo":null,"isNewFile":"1","line":"3","lineLength":"0","path":"tests\/test-bookflow.t","diffId":"10508","author":"pulkit","id":"68089","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"I am not good at bash I am certain you make it simpler by removing the logic related to pwd, I am not sure why do we need that.","replyTo":null,"isNewFile":"1","line":"2","lineLength":"0","path":"tests\/test-bookflow.t","diffId":"10508","author":"pulkit","id":"68088","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"output of push to \/dev\/null, that seems a bit weird, let's not do that.","replyTo":null,"isNewFile":"1","line":"29","lineLength":"0","path":"tests\/test-bookflow.t","diffId":"10508","author":"pulkit","id":"68087","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"If I understand correctly, the above couple of lines can be deleted?","replyTo":null,"isNewFile":"1","line":"89","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10508","author":"pulkit","id":"68086","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"We can better say that \"the active bookmark is updated to <commit>. use hg update to update to it.\"","replyTo":null,"isNewFile":"1","line":"72","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10508","author":"pulkit","id":"68085","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"We should move this couple of wrapfunctions call to uisetup() too.","replyTo":null,"isNewFile":"1","line":"87","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10508","author":"pulkit","id":"68084","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"This can be improved to \"creating named branches is disabled and you should use bookmarks. see `hg help bookmarks`.\"\n\nAlso maybe since we are defining a workflow using this extension, documenting that workflow will be great!","replyTo":null,"isNewFile":"1","line":"82","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10508","author":"pulkit","id":"68083","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"The `Run: hg up {}` part should be in the help text. Also `run 'hg up {}'` matches the core style more.","replyTo":null,"isNewFile":"1","line":"43","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10508","author":"pulkit","id":"68082","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"We generally use `cannot` instead of `can't` in core messages.","replyTo":null,"isNewFile":"1","line":"41","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10508","author":"pulkit","id":"68081","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"Can we switch the way help is described here. From 'cmdname: help' to\n\n`To create a new bookmark: hg book mark`.","replyTo":null,"isNewFile":"1","line":"14","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10508","author":"pulkit","id":"68080","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"Since the above two are enabled by default when we enabled this extension, they deserve a separate paragraph before the config options.","replyTo":null,"isNewFile":"1","line":"6","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10508","author":"pulkit","id":"68079","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"inline","comment":"Can we add some help text here that this extension helps implement feature branches workflow using bookamrks by providing config knobs to change the default behavior.","replyTo":null,"isNewFile":"1","line":"2","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10508","author":"pulkit","id":"68078","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"comment","comment":"I have looked at the things more thoroughly and commented things inline. Certain things can be treated as opinions and it's perfectly fine to now fix them.\n\nI found following cases missing in tests:\n  * Testing of creation of bookmark name which already exists\n  * Testing active bookmark is not moved on update (I might be wrong)\n\nAlso, mercurial core tests mainly test-check-code.t and test-check-commit.t fails with this patch. Please make them happy.\n\nAlso, the functionality this extension implemented is somewhat similar to https:\/\/www.mercurial-scm.org\/wiki\/BookmarkUpdatePlan.","author":"pulkit","id":"68077","dateCreated":"1535838664","dateModified":"1535838664"},{"type":"update","diffId":"10508","author":"idlsoft","id":"66848","dateCreated":"1534953419","dateModified":"1534953419"},{"type":"update","diffId":"10507","author":"idlsoft","id":"66847","dateCreated":"1534947382","dateModified":"1534947382"},{"type":"update","diffId":"10506","author":"idlsoft","id":"66846","dateCreated":"1534947146","dateModified":"1534947146"},{"type":"comment","comment":"I like the functionality it introduces and having this as an extension will be good.","author":"pulkit","id":"66823","dateCreated":"1534935242","dateModified":"1534935242"},{"type":"inline","comment":"please add r'' prefix to 'clean' and 'rev' for py3 compatibility.","replyTo":null,"isNewFile":"1","line":"70","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10406","author":"pulkit","id":"66822","dateCreated":"1534934965","dateModified":"1534934965"},{"type":"inline","comment":"You can enable the extension by putting `[extensions] bookflow=` in .hgrc. Aliasing does not looks neat.","replyTo":null,"isNewFile":"1","line":"2","lineLength":"0","path":"tests\/test-bookflow.t","diffId":"10406","author":"pulkit","id":"66821","dateCreated":"1534934965","dateModified":"1534934965"},{"type":"inline","comment":"We also don't allow double new lines in core.","replyTo":null,"isNewFile":"1","line":"53","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10406","author":"pulkit","id":"66820","dateCreated":"1534934965","dateModified":"1534934965"},{"type":"inline","comment":"_() missing here too.","replyTo":null,"isNewFile":"1","line":"71","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10406","author":"pulkit","id":"66819","dateCreated":"1534934965","dateModified":"1534934965"},{"type":"inline","comment":"_() missing here. Also, we don't start words with capital letters in error messages.","replyTo":null,"isNewFile":"1","line":"60","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10406","author":"pulkit","id":"66818","dateCreated":"1534934965","dateModified":"1534934965"},{"type":"inline","comment":"can we get rid of this hack and just use 'bookflow'?","replyTo":null,"isNewFile":"1","line":"25","lineLength":"0","path":"hgext\/bookflow.py","diffId":"10406","author":"pulkit","id":"66817","dateCreated":"1534934965","dateModified":"1534934965"},{"type":"update","diffId":"10406","author":"idlsoft","id":"66288","dateCreated":"1534461406","dateModified":"1534461406"}],"dateCreated":"1534461406","dateModified":"1537479138","status":"Needs Review"},{"id":"4621","callsign":"HG","title":"wireprotov2: teach changesetdata to fetch ancestors until depth","author":"indygreg","summary":"For shallow clone, it is useful to specify a starting node and tell\nthe server to send up to N ancestors from that starting point. This\nenables the server to perform the DAG walk without the client having\nto discover the base\/stop node(s) first.\n\nThis commit implements support for said queries on the changesetdata\ncommand.","testPlan":"","lineCount":"269","dependsOn":["4620"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11248","author":"indygreg","id":"71187","dateCreated":"1537478069","dateModified":"1537478069"},{"type":"accept","author":"durin42","id":"70663","dateCreated":"1537306641","dateModified":"1537306641"},{"type":"update","diffId":"11120","author":"indygreg","id":"70415","dateCreated":"1537219450","dateModified":"1537219450"}],"dateCreated":"1537219450","dateModified":"1537478069","status":"Accepted"},{"id":"4620","callsign":"HG","title":"wireprotov2: allow multiple fields to follow revision maps","author":"indygreg","summary":"The *data wire protocol commands emit a series of CBOR values.\nBecause revision\/delta data may be large, their data is emitted\noutside the map as a top-level bytestring value.\n\nBefore this commit, we'd emit a single optional bytestring\nvalue after the revision descriptor map. This got the job done.\nBut it was limiting in that we could only send a single field.\nAnd, it required the consumer to know that the presence of a\nkey in the map implied the existence of a following bytestring\nvalue.\n\nThis commit changes the encoding strategy so top-level bytestring\nvalues in the stream are explicitly denoted in a \"fieldsfollowing\"\nkey. This key contains an array defining what fields that follow\nand the expected size of each field.\n\nBy defining things this way, we can easily send N bytestring\nvalues without any ambiguity about their order. In addition,\nclients only need to know how to parse ``fieldsfollowing`` to\nknow if extra values are present.\n\nBecause this breaks backwards compatibility, we've bumped the version\nnumber of the wire protocol version 2 API endpoint.","testPlan":"","lineCount":"943","dependsOn":["4619"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11247","author":"indygreg","id":"71186","dateCreated":"1537478067","dateModified":"1537478067"},{"type":"update","diffId":"11119","author":"indygreg","id":"70405","dateCreated":"1537219447","dateModified":"1537219447"}],"dateCreated":"1537219447","dateModified":"1537478068","status":"Needs Review"},{"id":"4619","callsign":"HG","title":"wireprotov2: advertise set of valid values for requestable fields","author":"indygreg","summary":"changesetdata, manifestdata, and filedata all allow the caller to\nspecify what data fields to request.\n\nData fields are extensible and may evolve over time. In order to\nprevent clients from making requests for fields that are not\navailable, the client needs to know what fields are available.\n\nThis commit teaches the server to declare a set of \"valid values\"\nfor wire protocol command arguments. That set of values is exposed\nin the command's capabilities descriptor. The changesetdata,\nmanifestdata, and filedata commands all declare their set of\navailable \"fields.\"","testPlan":"","lineCount":"74","dependsOn":["4618"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11246","author":"indygreg","id":"71185","dateCreated":"1537478066","dateModified":"1537478066"},{"type":"accept","author":"durin42","id":"70661","dateCreated":"1537306505","dateModified":"1537306505"},{"type":"update","diffId":"11118","author":"indygreg","id":"70395","dateCreated":"1537219444","dateModified":"1537219444"}],"dateCreated":"1537219444","dateModified":"1537478066","status":"Accepted"},{"id":"4618","callsign":"HG","title":"wireprotov2: expose rich arguments metadata","author":"indygreg","summary":"Now that we internally store rich metadata about arguments, it makes\nsense to make that metadata available to the client. This will allow\nclients to validate outgoing command requests before they are sent\nover the wire.\n\nStrictly speaking, we should bump the wire protocol version for this\nchange since it is backwards incompatible. But no client-side code\ntouches the arguments map and I don't want to incur the work.","testPlan":"","lineCount":"305","dependsOn":["4616"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11245","author":"indygreg","id":"71184","dateCreated":"1537478065","dateModified":"1537478065"},{"type":"accept","author":"durin42","id":"70659","dateCreated":"1537306488","dateModified":"1537306488"},{"type":"update","diffId":"11117","author":"indygreg","id":"70385","dateCreated":"1537219442","dateModified":"1537219442"}],"dateCreated":"1537219442","dateModified":"1537478065","status":"Accepted"},{"id":"4616","callsign":"HG","title":"wireprotov2: advertise recognized path filter prefixes","author":"indygreg","summary":"While the wire protocol doesn't yet support it, we'll eventually\nhave commands that accept narrow patterns to specify the set of\nfiles relevant to a command.\n\nFor security and performance reasons, only specific filter types\nare allowed.\n\nThis commit teaches the server to advertise the set of allowed\nfilter types. By doing so, clients can e.g. validate user-specified\npatterns against the server's abilities without having to send\na command to retrieve data.\n\nHaving the data in the capabilities data structure will also serve\nas a check against unwanted BC.","testPlan":"","lineCount":"33","dependsOn":["4615"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11244","author":"indygreg","id":"71183","dateCreated":"1537478064","dateModified":"1537478064"},{"type":"accept","author":"durin42","id":"70655","dateCreated":"1537306394","dateModified":"1537306394"},{"type":"update","diffId":"11115","author":"indygreg","id":"70365","dateCreated":"1537219438","dateModified":"1537219438"}],"dateCreated":"1537219438","dateModified":"1537478064","status":"Accepted"},{"id":"4615","callsign":"HG","title":"wireprotov2: declare command arguments richly","author":"indygreg","summary":"Previously, we declared command arguments with an example of\ntheir value. After this commit, we declare command arguments\nas a dict of metadata. This allows us to define the value\ntype, whether the argument is required, and provide default\nvalues. This in turn allows us to have nice things, such as\nless boilerplate code in individual commands for validating\ninput and assigning default values. It should also make command\nbehavior more consistent as a result.\n\nTest output changed slightly because I realized that the \"fields\"\nargument wasn't being consistently defined as a set. Oops!\n\nOther test output changed because of slight differences in code\nperforming type validation.","testPlan":"","lineCount":"383","dependsOn":[],"reviewers":[],"ccs":["durin42","mercurial-devel"],"actions":[{"type":"update","diffId":"11243","author":"indygreg","id":"71182","dateCreated":"1537478063","dateModified":"1537478063"},{"type":"inline","comment":"We could do that, sure. Would it be OK if I made that change in D4617 instead?","replyTo":"70654","isNewFile":"1","line":"543","lineLength":"3","path":"mercurial\/wireprotov2server.py","diffId":"11114","author":"indygreg","id":"70666","dateCreated":"1537307439","dateModified":"1537307439"},{"type":"inline","comment":"Shouldn't there be a default XOR required == true? That is, could we get away with `default=None` and then deriving requiredness from default-availability?","replyTo":null,"isNewFile":"1","line":"543","lineLength":"3","path":"mercurial\/wireprotov2server.py","diffId":"11114","author":"durin42","id":"70654","dateCreated":"1537306345","dateModified":"1537306345"},{"type":"update","diffId":"11114","author":"indygreg","id":"70359","dateCreated":"1537219437","dateModified":"1537219437"}],"dateCreated":"1537219437","dateModified":"1537478063","status":"Needs Review"},{"id":"2679","callsign":"HG","title":"[PoC] obsolete: config option to enable local only obsolescence mode","author":"indygreg","summary":"Our path to enable obsolescence\/evolve in core is to enable\ncreation of markers locally (no exchange) with user-facing behavior\nthat mimics existing behavior as closely as possible. Think of it\nas evolve light.\n\nWe introduce a config option to control this behavior.","testPlan":"","lineCount":"14","dependsOn":[],"reviewers":[],"ccs":["durin42","lothiraldan","pulkit","mercurial-devel"],"actions":[{"type":"comment","comment":">>! In D2679#70934, @lothiraldan wrote:\n>>>! In D2679#69388, @durin42 wrote:\n>> I'm slowly becoming convinced that the long-unquestioned axiom that \"all markers are distributed globally\" isn't correct, and this is part of why: it's potentially of great value to be able to restore a change by re-pulling it, even though the obsmarkers would normally cause it to be deleted.\n> Avoiding silent reappearance of the locally obsolete changeset that we see on a remote repository is a core feature of obsolescence. Actually, it is the one issue that prompted the creation of changeset evolution in the first place. \n\nThere's a difference (in my view) between something accidentally coming back and intentionally coming back. The current state of affairs prevents both. I feel like we can do better.\n\n> We and the other people using distributed evolution rely on this behavior on a daily basis. The includes people who picked up the evolve extension on their own and came up with their own workflow without ever speaking to us. We can see some user interfaces adjustment in the future, but the set of synchronized markers and the associated behavior is something we are happy about. It has been well tested in diverse teams and situation for multiple years now.\n\nI've brought this up repeatedly and been dismissed with this exact argument every time. I weary of it, because it's not a new argument, it's just blind devotion to the initial design without any attempt at reflection if we can do better. I don't dispute that evolution as currently implemented is a good thing, but I feel like there's room for significant improvement while simultaneously simplifying the exchange algorithms. Care to try and convince me why I'm wrong, without anecdotal \"many people use this and haven't complained\" arguments?\n\n>> It's _super confusing_ when I `hg import` a patch and it seems to work but also immediately disappears, so I've got more sympathy for this PoC series than I do the theoretical purity of markers having any kind of globalness. Does that make sense?\n> This seems like an unrelated user interface issues. We usually warn the user when their working copy becomes obsolete, pointing out to the newest version\/evolution. We have to extend this messages logic to `hg import` to clarify the situation.\n\nI disagree, perhaps unsurprisingly.","author":"durin42","id":"70935","dateCreated":"1537377795","dateModified":"1537377795"},{"type":"comment","comment":">>! In D2679#69388, @durin42 wrote:\n>>>! In D2679#68489, @lothiraldan wrote:\n>> To take a step back, I'm wondering what's the end goal? I remember there was a discussion about having rebase enabled by default, is it related?\n> \n> Getting rebase (and maybe histedit?) enabled by default is my recollection of the rough goal.\n> \n>> \n>> The behavior target by this series (\"unobsolete\" re-pulled changeset) conflicts with the final behavior we want for Changeset Evolution. Intermediate steps are a good way to make progress. I feel like it is important to write down a clear plan when it comes to adding behavior that does not match our final goals. How are we planning to transition from the local-only step to full (ie, distributed) Evolution?\n> \n> I'm slowly becoming convinced that the long-unquestioned axiom that \"all markers are distributed globally\" isn't correct, and this is part of why: it's potentially of great value to be able to restore a change by re-pulling it, even though the obsmarkers would normally cause it to be deleted.\nAvoiding silent reappearance of the locally obsolete changeset that we see on a remote repository is a core feature of obsolescence. Actually, it is the one issue that prompted the creation of changeset evolution in the first place. We and the other people using distributed evolution rely on this behavior on a daily basis. The includes people who picked up the evolve extension on their own and came up with their own workflow without ever speaking to us. We can see some user interfaces adjustment in the future, but the set of synchronized markers and the associated behavior is something we are happy about. It has been well tested in diverse teams and situation for multiple years now.\n> It's _super confusing_ when I `hg import` a patch and it seems to work but also immediately disappears, so I've got more sympathy for this PoC series than I do the theoretical purity of markers having any kind of globalness. Does that make sense?\nThis seems like an unrelated user interface issues. We usually warn the user when their working copy becomes obsolete, pointing out to the newest version\/evolution. We have to extend this messages logic to `hg import` to clarify the situation.\n","author":"lothiraldan","id":"70934","dateCreated":"1537368619","dateModified":"1537368619"},{"type":"comment","comment":">>! In D2679#68489, @lothiraldan wrote:\n> To take a step back, I'm wondering what's the end goal? I remember there was a discussion about having rebase enabled by default, is it related?\n\nGetting rebase (and maybe histedit?) enabled by default is my recollection of the rough goal.\n\n> \n> The behavior target by this series (\"unobsolete\" re-pulled changeset) conflicts with the final behavior we want for Changeset Evolution. Intermediate steps are a good way to make progress. I feel like it is important to write down a clear plan when it comes to adding behavior that does not match our final goals. How are we planning to transition from the local-only step to full (ie, distributed) Evolution?\n\nI'm slowly becoming convinced that the long-unquestioned axiom that \"all markers are distributed globally\" isn't correct, and this is part of why: it's potentially of great value to be able to restore a change by re-pulling it, even though the obsmarkers would normally cause it to be deleted. It's _super confusing_ when I `hg import` a patch and it seems to work but also immediately disappears, so I've got more sympathy for this PoC series than I do the theoretical purity of markers having any kind of globalness. Does that make sense?","author":"durin42","id":"69389","dateCreated":"1536762660","dateModified":"1536762660"},{"type":"comment","comment":"Hi Pulkit,\n\nThanks for your patience. I had a couple of important things to get out of the way before I could turn my attention to this.\n\nI reached out to Ryan and he sent me the code he wrote at the 4.6 Sprint. I made it available here to support the discussion: https:\/\/phab.mercurial-scm.org\/D4479. The core differences in his approach are to filter out obsmarkers affecting the \"re-pulled\" nodes instead of \"stripping\" those markers. The main benefit is to avoid information loss. Stripping markers will break the evolution history, preventing user to benefit from it. Filtering markers out mean that the history will be fully restored if the affected changeset get obsolete again. It can be triggered by the same events.This is a pure implementation detail, having the same user-level effect than the behavior proposed in this Proof of Concept series.\n\nRegarding the implementation of this series, I feel like the current detection code needs to move at a higher level. Having changegroup aware of obsmarkers at such a low level seems wrong and can lead to consistency issues if the bundle contains  obsmarkers. An alternative approach would be to keep track of redundant incoming nodes during changegroup processing (on the transaction object as we are already doing for all new nodes), and process that list at the end of the transaction. We'll give a shot at such implementation, it is useful in other scenarios.\n\nTo take a step back, I'm wondering what's the end goal? I remember there was a discussion about having rebase enabled by default, is it related?\n\nThe behavior target by this series (\"unobsolete\" re-pulled changeset) conflicts with the final behavior we want for Changeset Evolution. Intermediate steps are a good way to make progress. I feel like it is important to write down a clear plan when it comes to adding behavior that does not match our final goals. How are we planning to transition from the local-only step to full (ie, distributed) Evolution?","author":"lothiraldan","id":"68489","dateCreated":"1536103480","dateModified":"1536103480"},{"type":"comment","comment":"Hey @lothiraldan, if not the patches itself, can you please explain the approach taken?\nIf I don't hear from you this week, I will push this series so that we can have enough time testing this in current cycle. Thanks!","author":"pulkit","id":"67297","dateCreated":"1535375938","dateModified":"1535375938"},{"type":"comment","comment":"@lothiraldan Gentle reminder for the series you were talking about.","author":"pulkit","id":"66824","dateCreated":"1534935417","dateModified":"1534935417"},{"type":"comment","comment":">>! In D2679#64079, @lothiraldan wrote:\n>>>! In D2679#62662, @pulkit wrote:\n>> These patches were result of a very extensive discussion about what ways we have to start supporting obsmarkers by default. I will like to push these changesets so that we can try these in this cycle. If I don't hear any concern in a week, I will push them.\n> \n> We made a series with Ryan during the sprint that had the advantage of not requiring to delete obs-markers. I will undust it, rebase it and send it here for making the discussion go forward.\n\nOh nice! That will be great! Looking forward to your series.","author":"pulkit","id":"64233","dateCreated":"1533639536","dateModified":"1533639536"},{"type":"comment","comment":">>! In D2679#62662, @pulkit wrote:\n> These patches were result of a very extensive discussion about what ways we have to start supporting obsmarkers by default. I will like to push these changesets so that we can try these in this cycle. If I don't hear any concern in a week, I will push them.\n\nWe made a series with Ryan during the sprint that had the advantage of not requiring to delete obs-markers. I will undust it, rebase it and send it here for making the discussion go forward.","author":"lothiraldan","id":"64080","dateCreated":"1533587987","dateModified":"1533587987"},{"type":"comment","comment":"These patches were result of a very extensive discussion about what ways we have to start supporting obsmarkers by default. I will like to push these changesets so that we can try these in this cycle. If I don't hear any concern in a week, I will push them.","author":"pulkit","id":"62663","dateCreated":"1533162260","dateModified":"1533162260"},{"type":"update","diffId":"6637","author":"indygreg","id":"43207","dateCreated":"1520210571","dateModified":"1520210571"}],"dateCreated":"1520210571","dateModified":"1537377795","status":"Needs Review"},{"id":"4664","callsign":"HG","title":"mergecommit: add a new extension to merge in-memory and create a commit","author":"pulkit","summary":"This patch adds a new mergecommit command which is very useful for workflows\nwhere there is automated merging involved. This patch uses existing in-memory\nmerge API to merge in-memory and create a commit.\n\nI was thinking about --commit flag to merge command, but then we need other\ncommitopts too, so created a new command.\n\nThis patches filemerge.py to not raise IMMConflictsError on first conflict\nbecause we need a list of conflicts in the end.","testPlan":"","lineCount":"108","dependsOn":[],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11198","author":"pulkit","id":"70929","dateCreated":"1537361525","dateModified":"1537361525"}],"dateCreated":"1537361525","dateModified":"1537361525","status":"Needs Review"},{"id":"4656","callsign":"HG","title":"revlog: move censor logic out of censor extension","author":"indygreg","summary":"The censor extension is doing very low-level things with revlogs.\nIt is fundamentally impossible for this logic to remain in the censor\nextension while support multiple storage backends: we need each\nstorage backend to implement censor in its own storage-specific\nway.\n\nThis commit effectively moves the revlog-specific censoring code to\nbe a method of revlogs themselves.\n\nWe've defined a new API on the file storage interface for censoring\nan individual node. Even though the current censoring code doesn't\nuse it, the API requires a transaction instance because it logically\nmakes sense for storage backends to require an active transaction\n(which implies a held write lock) in order to rewrite storage.\n\nAfter this commit, the censor extension has been reduced to\nboilerplate precondition checking before invoking the generic\nstorage API.\n\nI tried to keep the code as similar as possible. But some minor\nchanges were made:\n\n* We use self._io instead of instantiating a new revlogio instance.\n* We compare self.version against REVLOGV0 instead of != REVLOGV1\n  because presumably all future revlog versions will support censoring.\n* We use self.opener instead of going through repo.svfs (we don't have\n  a handle on the repo instance from a revlog).\n* \"revlog\" dropped\n* Replace \"flog\" with \"self\".","testPlan":"","lineCount":"201","dependsOn":["4655"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11190","author":"indygreg","id":"70850","dateCreated":"1537324445","dateModified":"1537324445"}],"dateCreated":"1537324445","dateModified":"1537324481","status":"Needs Review"},{"id":"4655","callsign":"HG","title":"global: replace most uses of RevlogError with StorageError (API)","author":"indygreg","summary":"When catching errors in storage, we should be catching\nStorageError instead of RevlogError. When throwing errors related\nto storage, we shouldn't be using RevlogError unless we know\nthe error stemmed from revlogs. And we only reliably know that\nif we're in revlog.py or are inheriting from a type defined in\nrevlog.py.","testPlan":"","lineCount":"54","dependsOn":["4654"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11189","author":"indygreg","id":"70840","dateCreated":"1537315563","dateModified":"1537315563"}],"dateCreated":"1537315563","dateModified":"1537324481","status":"Needs Review"},{"id":"4663","callsign":"HG","title":"filelog: stop proxying headrevs() (API)","author":"indygreg","summary":"The previous commit removed the last user of this method. It is\nredundant with heads() and adds little to no value other than\nconvenience. Let's nuke it.","testPlan":"","lineCount":"35","dependsOn":["4662"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11197","author":"indygreg","id":"70917","dateCreated":"1537324452","dateModified":"1537324452"}],"dateCreated":"1537324452","dateModified":"1537324452","status":"Needs Review"},{"id":"4662","callsign":"HG","title":"hgweb: use heads() instead of headrevs()","author":"indygreg","summary":"These appear to be the only callers of headrevs() on file storage\nobjects. Let's port to heads() so we can remove headrevs().","testPlan":"","lineCount":"4","dependsOn":["4661"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11196","author":"indygreg","id":"70907","dateCreated":"1537324451","dateModified":"1537324451"}],"dateCreated":"1537324451","dateModified":"1537324452","status":"Needs Review"},{"id":"4661","callsign":"HG","title":"filelog: record what's using attributes","author":"indygreg","summary":"filelog and the file storage interface have some attributes that\nideally shouldn't be there. This commit annotates some of those\nattributes with their users so we know where to look when it comes\ntime to removing them.\n\nThis exercise exposed a theme: many attributes are used by LFS,\nrepo upgrade, verify, and special repo types (like bundlerepo). That\npoints to missing abstractions on file storage to facilitate these\nspecial needs.","testPlan":"","lineCount":"22","dependsOn":["4660"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11195","author":"indygreg","id":"70897","dateCreated":"1537324450","dateModified":"1537324450"}],"dateCreated":"1537324450","dateModified":"1537324451","status":"Needs Review"},{"id":"4660","callsign":"HG","title":"filelog: stop proxying datafile (API)","author":"indygreg","summary":"It appears the censor code was the last user of this proxy. With\nthere being a dedicated censor API, we can drop the proxy.","testPlan":"","lineCount":"4","dependsOn":["4659"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11194","author":"indygreg","id":"70887","dateCreated":"1537324449","dateModified":"1537324449"}],"dateCreated":"1537324449","dateModified":"1537324450","status":"Needs Review"},{"id":"4659","callsign":"HG","title":"filelog: stop proxying _addrevision() (API)","author":"indygreg","summary":"There are no callers of this API in core. And I'm not sure why this\nproxy was added in the first place, as the commit that added it\n(1541e1a8e87d) didn't appear to have any callers in the repo either.\nWho knows.","testPlan":"","lineCount":"3","dependsOn":["4658"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11193","author":"indygreg","id":"70876","dateCreated":"1537324448","dateModified":"1537324448"}],"dateCreated":"1537324448","dateModified":"1537324449","status":"Needs Review"},{"id":"4658","callsign":"HG","title":"filelog: stop proxying compress() (API)","author":"indygreg","summary":"The censoring code was previously relying on this. With a dedicated\ncensoring API on the interface, no consumers are left and we can\nstop proxying this method.","testPlan":"","lineCount":"3","dependsOn":["4657"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11192","author":"indygreg","id":"70866","dateCreated":"1537324447","dateModified":"1537324447"}],"dateCreated":"1537324447","dateModified":"1537324448","status":"Needs Review"},{"id":"4657","callsign":"HG","title":"filelog: stop proxying start(), end(), and length() (API)","author":"indygreg","summary":"These were needed by the censoring code, which formerly lived in the\ncensor extension. Now that there is a censoring API on the file storage\ninterface, nothing uses these methods and we can stop proxying them.","testPlan":"","lineCount":"9","dependsOn":["4656"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11191","author":"indygreg","id":"70856","dateCreated":"1537324446","dateModified":"1537324446"}],"dateCreated":"1537324446","dateModified":"1537324447","status":"Needs Review"},{"id":"4654","callsign":"HG","title":"error: introduce StorageError","author":"indygreg","summary":"Errors in revlogs are often represented by RevlogError. It's fine\nfor revlogs to raise a revlog-specific exception. But in the context\nof multiple storage backends, it doesn't make sense to be throwing or\ncatching an exception with \"revlog\" in its name when revlogs may not\neven be in play.\n\nThis commit introduces a new generic StorageError type for representing\nerrors in the storage layer.\n\nRevlogError is an instance of this type.\n\nInterface documentation and tests referencing RevlogError has been\nupdated to specify StorageError should be used.\n\n.. api::\n\n   ``error.StorageError`` has been introduced to represent errors in\n   storage. It should be used in place of ``error.RevlogError`` unless\n   the error is known to come from a revlog.","testPlan":"","lineCount":"31","dependsOn":["4653"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11188","author":"indygreg","id":"70830","dateCreated":"1537315561","dateModified":"1537315561"}],"dateCreated":"1537315561","dateModified":"1537315563","status":"Needs Review"},{"id":"4653","callsign":"HG","title":"revlog: drop LookupError alias (API)","author":"indygreg","summary":"This alias is especially bad because it shadows the built-in\nLookupError type. This has caused me confusion in the past\nwhen reading revlog code. Qualifying all uses with \"error.\" will\nmake it obvious that we're using a Mercurial error type.","testPlan":"","lineCount":"24","dependsOn":["4652"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11187","author":"indygreg","id":"70820","dateCreated":"1537315560","dateModified":"1537315560"}],"dateCreated":"1537315560","dateModified":"1537315561","status":"Needs Review"},{"id":"4652","callsign":"HG","title":"revlog: drop some more error aliases (API)","author":"indygreg","summary":"These should be lightly used and I doubt that will be any\nstrong objections to removing the aliases.\n\nNote that some uses of ProgrammingError in this file use\ntranslated messages. I'm pretty sure that's a bug. But the\nlinters don't complain, so meh.","testPlan":"","lineCount":"24","dependsOn":["4651"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11186","author":"indygreg","id":"70810","dateCreated":"1537315559","dateModified":"1537315559"}],"dateCreated":"1537315559","dateModified":"1537315560","status":"Needs Review"},{"id":"4651","callsign":"HG","title":"revlog: drop RevlogError alias (API)","author":"indygreg","summary":"error.RevlogError was moved from revlog.py in 08cabecfa8a8 in\n2009. revlog.RevlogError has remained as an alias ever since.\n\nLet's drop the alias and use error.RevlogError directly.","testPlan":"","lineCount":"84","dependsOn":["4650"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11185","author":"indygreg","id":"70799","dateCreated":"1537315558","dateModified":"1537315558"}],"dateCreated":"1537315558","dateModified":"1537315559","status":"Needs Review"},{"id":"4650","callsign":"HG","title":"testing: add interface unit tests for file storage","author":"indygreg","summary":"Our strategy for supporting alternate storage backends is to define\ninterfaces for everything then \"code to the interface.\"\n\nWe already have interfaces for various primitives, including file\nand manifest storage.\n\nWhat we don't have is generic unit tests for those interfaces. Up\nto this point we've been relying on high-level integration tests\n(mainly in the form of existing .t tests) to test alternate storage\nbackends. And my experience with developing the \"simple store\" test\nextension is that such testing is very tedious: it takes several\nminutes to run all tests and when you find a failure, it is often\nnon-trivial to debug.\n\nThis commit starts to change that.\n\nThis commit introduces the mercurial.testing.storage module. It\ncontains testing code for storage. Currently, it defines some\nunittest.TestCase classes for testing the file storage interfaces.\nIt also defines some factory functions that allow a caller to easily\nspawn a custom TestCase \"bound\" to a specific file storage backend\nimplementation.\n\nA new .py test has been added. It simply defines a callable to produce\nfilelog and transaction instances on demand and then \"registers\" the\nvarious test classes so the filelog class can be tested with the\nstorage interface unit tests.\n\nAs part of writing the tests, I identified a couple of apparent\nbugs in revlog.py and filelog.py! These are tracked with inline\nTODO comments.\n\nWriting the tests makes it more obvious where the storage interface\nis lacking. For example, we raise either IndexError or\nerror.LookupError for missing revisions depending on whether we\nuse an integer revision or a node. Also, we raise error.RevlogError\nin various places when we should be raising a storage-agnostic\nerror type.\n\nThe storage interfaces are currently far from perfect and there is much\nwork to be done to improve them. But at least with this commit we\nfinally have the start of unit tests that can be used to \"qualify\"\nthe behavior of a storage backend. And when implementing and debugging\nnew storage backends, we now have an obvious place to define new\ntests and have obvious places to insert breakpoints to facilitate\ndebugging. This should be invaluable when implementing new storage\nbackends.\n\nI added the mercurial.testing package because these interface\nconformance tests are generic and need to be usable by all storage\nbackends. Having the code live in tests\/ would make it difficult for\nstorage backends implemented in extensions to test their interface\nconformance. First, it would require obtaining a copy of Mercurial's\nstorage test code in order to test. Second, it would make testing\nagainst multiple Mercurial versions difficult, as you would need to\nimport N copies of the storage testing code in order to achieve test\ncoverage. By making the test code part of the Mercurial distribution\nitself, extensions can `import mercurial.testing.*` to access and run\nthe test code. The test will run against whatever Mercurial version\nis active.\n\nFWIW I've always wanted to move parts of run-tests.py into the\nmercurial.* package to make the testing story simpler (e.g. imagine an\n`hg debugruntests` command that could invoke the test harness). While I\nhave no plans to do that in the near future, establishing the\nmercurial.testing package does provide a natural home for that code\nshould someone do this in the future.","testPlan":"","lineCount":"1031","dependsOn":[],"reviewers":[],"ccs":["mjpieters","mercurial-devel"],"actions":[{"type":"update","diffId":"11184","author":"indygreg","id":"70793","dateCreated":"1537315556","dateModified":"1537315556"}],"dateCreated":"1537315556","dateModified":"1537315558","status":"Needs Review"},{"id":"4649","callsign":"HG","title":"narrow: remove narrowrevlog","author":"indygreg","summary":"Core now automatically enables ellipsis support on revlogs when\nrepositories have narrow enabled. So, we no longer need to globally\nregister the revlog flag as part of activating the narrow extension\nand this code can be deleted.\n\nA side effect of this change is that repositories will now raise an\nerror on encountering an ellipsis flag when the narrow extension is\nloaded. Previously, loading the narrow extension on a non-narrow repo\ncould result in silent usage of the ellipsis flag. This could lead\nto undetected bugs. I think the new behavior is more correct.","testPlan":"","lineCount":"30","dependsOn":["4648"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11183","author":"indygreg","id":"70779","dateCreated":"1537310447","dateModified":"1537310447"}],"dateCreated":"1537310447","dateModified":"1537310447","status":"Needs Review"},{"id":"4648","callsign":"HG","title":"localrepo: enable ellipsis flag on revlogs when repo is narrow","author":"indygreg","summary":"If the narrow requirement is present, revlogs created for that\nrepository will have the ellipsis flag enabled.\n\nThis is the same behavior that the narrow extension exhibits. Except\nthe ellipsis flag won't be enabled on repos\/revlogs that don't have\nthe narrow requirement.","testPlan":"","lineCount":"3","dependsOn":["4647"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11182","author":"indygreg","id":"70769","dateCreated":"1537310444","dateModified":"1537310444"}],"dateCreated":"1537310444","dateModified":"1537310447","status":"Needs Review"},{"id":"4647","callsign":"HG","title":"revlog: add opener option to enable ellipsis flag processor","author":"indygreg","summary":"The ellipsis flag processor can now be registered by specifying\nan opener option when constructing a revlog instance. This allows\nus to enable ellipsis flags on a per-revlog basis.","testPlan":"","lineCount":"2","dependsOn":["4646"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11181","author":"indygreg","id":"70759","dateCreated":"1537310442","dateModified":"1537310442"}],"dateCreated":"1537310442","dateModified":"1537310444","status":"Needs Review"},{"id":"4646","callsign":"HG","title":"revlog: store flag processors per revlog","author":"indygreg","summary":"Previously, revlog flag processing would consult a global dict\nwhen processing flags. This was simple. But it had the undesired\nside-effect that any extension could load flag processors once\nand those flag processors would be available to any revlog that was\nsubsequent loaded in the process. e.g. in hgweb, if the narrow\nextension were loaded for repo A but not repo B, repo B would be\nable to decode ellipsis flags even though it shouldn't be able to.\n\nMaking the flag processors dict per-revlog allows us to have per-revlog\ncontrols over what flag processors are available, thus preserving\ndesired granular access to flag processors depending on the revlog's\nneeds.\n\nIf a flag processor is globally registered, it is still globally\navailable. So this commit should not meaningfully change behavior.","testPlan":"","lineCount":"8","dependsOn":["4645"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11180","author":"indygreg","id":"70749","dateCreated":"1537310439","dateModified":"1537310439"}],"dateCreated":"1537310439","dateModified":"1537310442","status":"Needs Review"},{"id":"4645","callsign":"HG","title":"revlog: define ellipsis flag processors in core","author":"indygreg","summary":"We will soon be teaching core to honor the ellipsis flag on revlogs.\nMoving the definition of the processor functions to core is the first\nstep in this.\n\nThe processor is still not registered unless the narrow extension is\nloaded.","testPlan":"","lineCount":"28","dependsOn":["4644"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11179","author":"indygreg","id":"70738","dateCreated":"1537310437","dateModified":"1537310437"}],"dateCreated":"1537310437","dateModified":"1537310439","status":"Needs Review"},{"id":"4644","callsign":"HG","title":"narrow: remove custom filelog type","author":"indygreg","summary":"This functionality is now handled by core as of the previous commit.\n\nI wanted this to be a standalone commit because the deleted code\nmakes a reference to remotefilelog's file type missing a node() method\nand this may have implications to narrow+remotefilelog usage. The code\nin core doesn't perform this check and therefore behavior may be subtly\ndifferent and buggy.\n\nBut I \/think\/ the check is merely a performance optimization and\nnothing more. So I'm optimistic this will continue to \"just work.\"","testPlan":"","lineCount":"57","dependsOn":["4643"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11178","author":"indygreg","id":"70727","dateCreated":"1537310435","dateModified":"1537310435"}],"dateCreated":"1537310435","dateModified":"1537310437","status":"Needs Review"},{"id":"4643","callsign":"HG","title":"filelog: custom filelog to be used with narrow repos","author":"indygreg","summary":"Narrow repos may have file revisions whose copy\/rename metadata\nreferences files not in the store. This can pose problems when\nconsumers attempt to access a missing referenced file revision.\n\nThe narrow extension hacks around this problem by implementing a\nderived filelog type that provides custom implementations of\nrenamed(), size(), and cmp() which handle renames against files not\nin the narrow spec by silently removing the rename metadata.\n\nWhile silently dropping metadata isn't the most robust solution,\nit is the easiest to implement.\n\nThis commit ports the custom narrow filelog class to core.\n\nWhen a narrow repo is constructed, its ifilestorage creation\nfunction will automatically use the new filelog type. This means\nthe extra logic is 0 cost for non-narrow repos and shouldn't\ninterfere with their operation.","testPlan":"","lineCount":"66","dependsOn":["4642"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11177","author":"indygreg","id":"70717","dateCreated":"1537310433","dateModified":"1537310433"}],"dateCreated":"1537310433","dateModified":"1537310435","status":"Needs Review"},{"id":"4642","callsign":"HG","title":"localrepo: iteratively derive local repository type","author":"indygreg","summary":"This commit implements the dynamic local repository type derivation\nthat was explained in the recent commit\nbfeab472e3c0 \"localrepo: create new function for instantiating a local\nrepo object.\"\n\nInstead of a static localrepository class\/type which must be customized\nafter construction, we now dynamically construct a type by building up\nbase classes\/types to represent specific repository interfaces.\n\nConceptually, the end state is similar to what was happening when\nvarious extensions would monkeypatch the __class__ of newly-constructed\nrepo instances. However, the approach is inverted. Instead of making\nthe instance then customizing it, we do the customization up front\nby influencing the behavior of the type then we instantiate that\ncustom type.\n\nThis approach gives us much more flexibility. For example, we can\nuse completely separate classes for implementing different aspects\nof the repository. For example, we could have one class representing\nrevlog-based file storage and another representing non-revlog based\nfile storage. When then choose which implementation to use based on\nthe presence of repo requirements.\n\nA concern with this approach is that it creates a lot more types\nand complexity and that complexity adds overhead. Yes, it is true that\nthis approach will result in more types being created. Yes, this is\nmore complicated than traditional \"instantiate a static type.\" However,\nI believe the alternatives to supporting alternate storage backends\nare just as complicated. (Before I arrived at this solution, I had\npatches storing factory functions on local repo instances for e.g.\nconstructing a file storage instance. We ended up having a handful\nof these. And this was logically identical to assigning custom\nmethods. Since we were logically changing the type of the instance,\nI figured it would be better to just use specialized types instead\nof introducing levels of abstraction at run-time.)\n\nOn the performance front, I don't believe that having N base classes\nhas any significant performance overhead compared to just a single base\nclass. Intuition says that Python will need to iterate the base classes\nto find an attribute. However, CPython caches method lookups: as long as\nthe __class__ or MRO isn't changing, method attribute lookup should be\nconstant time after first access. And non-method attributes are stored\nin __dict__, of which there is only 1 per object, so the number of\nbase classes for __dict__ is irrelevant.\n\nAnyway, this commit splits up the monolithic completelocalrepository\ninterface into sub-interfaces: 1 for file storage and 1 representing\neverything else.\n\nWe've taught ``makelocalrepository()`` to call a series of factory\nfunctions which will produce types implementing specific interfaces.\nIt then calls type() to create a new type from the built-up list of\nbase types.\n\nThis commit should be considered a start and not the end state. I\nsuspect we'll hit a number of problems as we start to implement\nalternate storage backends:\n\n* Passing custom arguments to __init__ and setting custom attributes\n  on __dict__.\n* Customizing the set of interfaces that are needed. e.g. the\n  \"readonly\" intent could translate to not requesting an interface\n  providing methods related to writing.\n* More ergonomic way for extensions to insert themselves so their\n  callbacks aren't unconditionally called.\n* Wanting to modify vfs instances, other arguments passed to __init__.\n\nThat being said, this code is usable in its current state and I'm\nconvinced future commits will demonstrate the value in this approach.","testPlan":"","lineCount":"149","dependsOn":["4641"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11176","author":"indygreg","id":"70706","dateCreated":"1537310431","dateModified":"1537310431"}],"dateCreated":"1537310431","dateModified":"1537310433","status":"Needs Review"},{"id":"4641","callsign":"HG","title":"localrepo: pass root manifest into manifestlog.__init__","author":"indygreg","summary":"Today, localrepository has a method that can be overloaded which\nreturns an instance of the root manifest storage object. When a\nmanifestlog is created, it calls this private method and stores\nthe root manifest object on it.\n\nThis \"hook\" on localrepository isn't part of the documented interface.\nIt isn't compatible with our desire to make repo storage determined\nbefore the repo object is constructed.\n\nThis commit changes manifestlog.__init__ to accept the root\nstorage object instead of calling into the repo to construct it.\nBy doing things this way, the repo instance is responsible for\nconstructing the manifest storage object directly.\n\nThis does mean that other derived repo types need to overload\nmanifestlog(). But they should have been doing this already,\nas manifestlog() is typically decorated in a storage-specific way.\ne.g. localrepository.manifestlog() is decorated as\n@storecache('00manifest.i'). And this assumes that a 00manifest.i\nfile exists in the store vfs. This condition may not hold for\nrepository types using non-revlog storage. So it is important\nfor special repo types to override manifestlog() to remove this\nfile association.\n\nThe code changed in perf is wrong because it isn't compatible with\nolder Mercurial versions. But I'm pretty sure the code was broken\non older versions before this commit. It only affects `hg perftags`.\nI don't care enough to fix that at this time.\n\n.. api::\n\n   ``manifest.manifestlog.__init__()`` now receives the root manifest\n   storage instance instead of calling into a private method on\n   the repo object to obtain it.","testPlan":"","lineCount":"36","dependsOn":[],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"11175","author":"indygreg","id":"70700","dateCreated":"1537310430","dateModified":"1537310430"}],"dateCreated":"1537310430","dateModified":"1537310431","status":"Needs Review"},{"id":"4479","callsign":"HG","title":"[RFC] obsolete: add mechanism to filter obsstore","author":"lothiraldan","summary":"This patch is not ready to land and only sent for discussion.","testPlan":"","lineCount":"24","dependsOn":[],"reviewers":[],"ccs":["pulkit","martinvonz","mercurial-devel"],"actions":[{"type":"comment","comment":"I just had a quick look. obsmarker repr hash, are we kind of inventing a way to index obsmarkers? I am pretty sure you people have thought about that, can you explain why you went with obsmarker repr hash instead of including a real index?\n\nAlso, feel free to point me to old discussions if this was discussed earlier.","author":"pulkit","id":"68685","dateCreated":"1536183949","dateModified":"1536183949"},{"type":"comment","comment":"I'll reply to my own question: The context is in D2679.","author":"martinvonz","id":"68491","dateCreated":"1536103749","dateModified":"1536103749"},{"type":"comment","comment":"Why? (Sorry if I'm asking to early and you were just about to send an update.)","author":"martinvonz","id":"68488","dateCreated":"1536103366","dateModified":"1536103366"},{"type":"update","diffId":"10777","author":"lothiraldan","id":"68482","dateCreated":"1536103193","dateModified":"1536103193"}],"dateCreated":"1536103193","dateModified":"1536183949","status":"Needs Review"},{"id":"4409","callsign":"HG","title":"rebase: skip *all* obsolete revisions, just warn about divergence","author":"martinvonz","summary":"We already skip obsolete revisions in some case: if there is no\nsuccessor, or when the successor is in the rebase set or in the\ndestination, I think. However, we instead error out if the successor\nis elsewhere (e.g. a child or a sibling of the destination), because\nit would cause divergence. The previous commit fixed one specific case\nof this.\n\nConsider this history (based on a test case updated by this patch):\n\n  o  D\n  |\n  | o  B'\n  |\/\n  | o  C\n  | |\n  | x  B\n  |\/\n  o A\n\nIf the user now runs `hg rebase -s B -d D` we would error out and tell\nthem that it would cause divergence from B. This patch makes it so we\ninstead ignore B and C (and tell the user that we're doing that\nbecause it would cause divergence). I think this simpler model will be\neasier for users to understand (I had also thought that all obsolete\ncommits and their descendants were skipped until we got reports from\nusers that that wasn't the case).\n\nNote that we still skip just the obsolete (not its descendants) if\nit's a prune or if the successor is in the destination.\n\nThis should also let us simplify the source quite a bit. I'll do that\nin later commits after I've heard people's thoughts on this one.","testPlan":"","lineCount":"40","dependsOn":[],"reviewers":[],"ccs":["lothiraldan","yuja","mercurial-devel"],"actions":[{"type":"comment","comment":"Changing the behavior of `--src` to ignore obsolete and orphan unrelated to the destination seems like a good idea.\n\nHowever, this change seems to also affect `--rev`. The `--rev` allow for a precise selection of what the user wants to see rebased and it feels like we should stick to the user wishes here. Dropping explicitly selected revisions feels wrong and will make some use cases harder to perform.","author":"lothiraldan","id":"67654","dateCreated":"1535551219","dateModified":"1535551219"},{"type":"comment","comment":"> @@ -1289,18 +1286,16 @@\n>    o  0:b173517d0057 a\n>    \n>    $ hg rebase -d 0 -r 2\n> -  rebasing 2:a82ac2b38757 \"c\" (c)\n> +  note: not rebasing 2:a82ac2b38757 \"c\" (c) and its descendants as this would cause divergence\n\nLooks like the fix for issue5782 regressed.\n\nhttps:\/\/bz.mercurial-scm.org\/show_bug.cgi?id=5782","author":"yuja","id":"67642","dateCreated":"1535547342","dateModified":"1535547342"},{"type":"update","diffId":"10616","author":"martinvonz","id":"67528","dateCreated":"1535441604","dateModified":"1535441604"}],"dateCreated":"1535441604","dateModified":"1535551219","status":"Needs Review"},{"id":"4283","callsign":"HG","title":"webcommands: fix `@webcommand` decorator","author":"sheehan","summary":"This commit fixes the `@webcommand` decorator used to define\nwebcommands. Although webcommands are currently using this\ndecorator, it does not behave as intended. Using the decorator\nshould record the name of the decorator in a `commands` dict,\nand when a request arrives to hgweb the correct function to\ncall should be resolved from that dict.\n\nIn the current implementation the command is added to the\ncommands dict, but when hgweb attempts to resolve the function\nwhile processing a request, it instead looks for an attribute\nwith the same name attached to the `webcommands` module. To\nsummarize, a command 'commandname' should be resolved from\n\n    `mercurial.hgweb.webcommands.commands['commandname']`\n\nbut is instead being resolved from\n\n    `mercurial.hgweb.webcommands.commandname`\n\nThe decorator appears to be working on the core webcommands\n(such as log, rev, file) however this is because those\ncommands are defined with the same name in the underlying\nPython function.\n\nThis commit changes hgweb_mod.py to resolve the function used\nfrom webcommands to come from the `commands` dict in the\n`webcommands` module. Due to the change in how webcommands\nare resolved, a `wrapwebcommand` function is also added, which\nwraps a webcommand by applying the wrapper as a partial function\nand overwrites the existing function in the `commands` dict.\n\nWrapped webcommands in shipped extensions and hgweb tests\nare changed to use the fixed decorator and `wrapwebcommand`\nfunction.","testPlan":"","lineCount":"48","dependsOn":[],"reviewers":[],"ccs":["foozy","indygreg","mercurial-devel"],"actions":[{"type":"comment","comment":"Sorry, I overlooked delivery error of reply to phabricator.\n\nThis is re-sending of the reply at Sat, 18 Aug 2018 15:18:10 +0900,\nwhich was cc-ed to mercurial-devel@mercurial-scm.org.\n\nAt Wed, 15 Aug 2018 19:00:51 +0000,\nsheehan (Connor Sheehan) wrote:\n[snip]\n> diff --git a\/mercurial\/hgweb\/hgweb_mod.py b\/mercurial\/hgweb\/hgweb_mod.py\n> --- a\/mercurial\/hgweb\/hgweb_mod.py\n> +++ b\/mercurial\/hgweb\/hgweb_mod.py\n> @@ -354,7 +354,7 @@\n>                  cmd = cmd[style + 1:]\n>  \n>              # avoid accepting e.g. style parameter as command\n> -            if util.safehasattr(webcommands, cmd):\n> +            if cmd in webcommands.commands:\n>                  req.qsparams['cmd'] = cmd\n>  \n>              if cmd == 'static':\n> @@ -416,15 +416,15 @@\n>  \n>                  res.headers['ETag'] = tag\n>  \n> -            if cmd not in webcommands.__all__:\n> +            if cmd not in webcommands.commands:\n>                  msg = 'no such method: %s' % cmd\n>                  raise ErrorResponse(HTTP_BAD_REQUEST, msg)\n>              else:\n>                  # Set some globals appropriate for web handlers. Commands can\n>                  # override easily enough.\n>                  res.status = '200 Script output follows'\n>                  res.headers['Content-Type'] = ctype\n> -                return getattr(webcommands, cmd)(rctx)\n> +                return webcommands.commands[cmd](rctx)\n>  \n>          except (error.LookupError, error.RepoLookupError) as err:\n>              msg = pycompat.bytestr(err)\n\nIMHO, for backward compatibility, this kind of (API) change should\nhave deprecation period before stopping support, and should use\nui.deprecwan() while such period, like below (this implementation\nassumes introduction of new decorator registrar.webcommand).\n\n    def _getwebcommand(ui, name):\n        cmd = webcommands.commands.get(name, None)\n        if not cmd:\n            if (name in webcommands.__all__ and\n                util.safehasattr(webcommands, name)):\n                cmd = getattr(webcommands, name)\n                # store into webcommands.commands for subsequent access\n                webcommands.commands[name] = cmd\n\n                ui.deprecwarn('legacy registration for webcommand %s,'\n                              ' use registrar.webcommand decorator' % name,\n                              '5.0')\n        return cmd\n\n\n> diff --git a\/hgext\/largefiles\/uisetup.py b\/hgext\/largefiles\/uisetup.py\n> --- a\/hgext\/largefiles\/uisetup.py\n> +++ b\/hgext\/largefiles\/uisetup.py\n> @@ -151,7 +151,7 @@\n>      extensions.wrapfunction(archival, 'archive', overrides.overridearchive)\n>      extensions.wrapfunction(subrepo.hgsubrepo, 'archive',\n>                              overrides.hgsubrepoarchive)\n> -    extensions.wrapfunction(webcommands, 'archive', overrides.hgwebarchive)\n> +    webcommands.wrapwebcommand('archive', overrides.hgwebarchive)\n>      extensions.wrapfunction(cmdutil, 'bailifchanged',\n>                              overrides.overridebailifchanged)\n\nAfter changes above, direct invocation of original webcommand function\n'webcommands.archive()' from another code path avoids wrapping effect,\nbecause 'wrapwebcommand' wraps only the function stored in dispatch\ntable 'webcommands.command'.\n\nNot for webcommand wrapping example, but for normal command wrapping\nexample, largefiles extension wraps both 'rebase' command and 'rebase'\nfunction. In largefiles\/uisetup.py::\n\n        if name == 'rebase':\n            extensions.wrapcommand(getattr(module, 'cmdtable'), 'rebase',\n                overrides.overriderebase)\n            extensions.wrapfunction(module, 'rebase',\n                                    overrides.overriderebase)\n\nOf course, it will be rare that un-wrapped original function causes\nproblem in webcommand case. But \"wrap function in dispatch table\"\nshould be separated from \"stop wrapping original webcommand function\"\nat first, IMHO.\n\n\nI'm also working around webcommand registration, and will post patches\nsoon. But, of course, feel free to re-send your revised one, for\nfeedback between us :-)","author":"foozy","id":"66813","dateCreated":"1534900316","dateModified":"1534900316"},{"type":"comment","comment":">>! In D4283#65988, @indygreg wrote:\n> This seems like a strict improvement.\n> \n> But the proper way to register web commands from extensions would be to go through the `registrar` API and have the extension loader look for a well-named symbol in each extension module that is loaded and hgweb would consult the registrar for active commands. In theory, this will only activate web commands on repositories that have an extension loaded.\n> \n> Search for `templatefilter` in `mercurial\/extensions.py` for an example of how all this works.\n> \n> Would you be willing to try that approach? It doesn't have to be perfect. But we are moving to the registrar for extensions wishing to install well-defined things. And web commands fit that bill.\n\nSeems reasonable to me! I took a look through the code and I think I understand what is going on. I'll update this patch with something that uses the registrar API.","author":"sheehan","id":"66593","dateCreated":"1534772811","dateModified":"1534772811"},{"type":"plan-changes","author":"sheehan","id":"66592","dateCreated":"1534772811","dateModified":"1534772811"},{"type":"comment","comment":"This seems like a strict improvement.\n\nBut the proper way to register web commands from extensions would be to go through the `registrar` API and have the extension loader look for a well-named symbol in each extension module that is loaded and hgweb would consult the registrar for active commands. In theory, this will only activate web commands on repositories that have an extension loaded.\n\nSearch for `templatefilter` in `mercurial\/extensions.py` for an example of how all this works.\n\nWould you be willing to try that approach? It doesn't have to be perfect. But we are moving to the registrar for extensions wishing to install well-defined things. And web commands fit that bill.","author":"indygreg","id":"65989","dateCreated":"1534378968","dateModified":"1534378968"},{"type":"update","diffId":"10360","author":"sheehan","id":"65975","dateCreated":"1534359638","dateModified":"1534359638"}],"dateCreated":"1534359638","dateModified":"1534900316","status":"Changes Planned"},{"id":"4301","callsign":"HG","title":"tests: add a directory 'knownfailures' to check in known failures","author":"pulkit","summary":"There are some things which everyone of us stumble upon and found out that this\ndoes not work. Sometimes, we hg developers know that these things does not works\nbecause of experience. But this does not scale well when we have a new developer\nor maybe a user who wants to see what does not work and is interested in fixing\nthat.\n\nChecking in the tests about the things which don't work will help others to\nreproduce these and also helps in fixing because of functionalities\n.\/run-tests.py provides.\n\nIIRC we had this thought in one of the previous sprints also to checkin some of\nthe known failures.\n\nI am starting with checking in a basic test failure. But I plan to check in\nmore narrow related failures which we have found.\n\nUnable to find easily what exactly broken with merges and narrow made me to\nimplement this idea so that we won't have similar problem in future.","testPlan":"","lineCount":"4","dependsOn":[],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"comment","comment":"I also think that the 'knowfailures' can serve as a good place for new contributors.\nI don't want to check-in all the failures and leave them unfixed. It's like I don't know which one I will be able to fix and I have a reproduction which can help others.","author":"pulkit","id":"66188","dateCreated":"1534432054","dateModified":"1534432054"},{"type":"update","diffId":"10391","author":"pulkit","id":"66182","dateCreated":"1534431586","dateModified":"1534431586"}],"dateCreated":"1534431586","dateModified":"1534432054","status":"Needs Review"},{"id":"3896","callsign":"HG","title":"copies: handle case when both merge cset are not descendant of merge base","author":"pulkit","summary":"There can be cases when both the changesets which we are merging are not\ndescendants of the merge base. In those cases dirtyc1 and dirtyc2 both will be\ntrue. The existing code assumes that either of them will be true always but that\nis not a right assumption.\n\nI found this while working with content-divergence resolution. In\ncontent-divergence resolution, we use the common predecessor as the merge base\nfor resolving content divergence and there it can be possible that the merge\nbase is not the descendant of both the content-divergence changsets.\n\nThe code in content-divergence which does this is at:\nhttps:\/\/www.mercurial-scm.org\/repo\/evolve\/file\/b81fd1487e04\/hgext3rd\/evolve\/evolvecmd.py#l507","testPlan":"","lineCount":"5","dependsOn":[],"reviewers":[],"ccs":["spectral","yuja","mercurial-devel"],"actions":[{"type":"comment","comment":">>! In D3896#65986, @spectral wrote:\n> Any updates on this?  I'm poking at updating this function in core, and not too excited to have to also modify the copy in evolve that seems to have spawned :\/\n\nI am yet to write tests. Feel free to update the one in core, I will take care of the evolve copy.","author":"pulkit","id":"66180","dateCreated":"1534428512","dateModified":"1534428512"},{"type":"comment","comment":"Any updates on this?  I'm poking at updating this function in core, and not too excited to have to also modify the copy in evolve that seems to have spawned :\/","author":"spectral","id":"65987","dateCreated":"1534370496","dateModified":"1534370496"},{"type":"comment","comment":"Can you add some tests that make both `dirtyc1` and `dirtyc2` set, and\ntrigger copy tracing?\n\n> Since we can have both dirtyc1, and dirtyc2 true, I am not sure whether\n> this else should be turned into it's own if statement?\n\nNo idea. I doubt if the current copy tracing can handle such cases. From\nmy vague memory, the current algorithm relies on the fact that the pseudo\n`base` revision exists in between the `c1` and the `c2`.","author":"yuja","id":"60811","dateCreated":"1531138991","dateModified":"1531138991"},{"type":"inline","comment":"Since we can have both dirtyc1, and dirtyc2 true, I am not sure whether this else should be turned into it's own if statement?","replyTo":null,"isNewFile":"1","line":"496","lineLength":"0","path":"mercurial\/copies.py","diffId":"9471","author":"pulkit","id":"60802","dateCreated":"1531041256","dateModified":"1531041256"},{"type":"update","diffId":"9471","author":"pulkit","id":"60797","dateCreated":"1531041116","dateModified":"1531041116"}],"dateCreated":"1531041116","dateModified":"1534428512","status":"Needs Review"},{"id":"4053","callsign":"HG","title":"[RFC]hgweb: garbage collect on every request in hgweb_mod too","author":"pulkit","summary":"We recently updated to mercurial 4.6.2 on our servers and also increased number\nof requests to be made on server at one time. This made hgweb instance takes up\na lot of memory, sometime upto 50GB of memory when requests are made parallely.\n\nThis patch is motivated from\nhttps:\/\/www.mercurial-scm.org\/repo\/hg-committed\/rev\/ff2370a70fe8 which adds\ngc.collect() call for the server which serves multiple repositories.\n\nI tried profiling this fix, sometimes this shows less memory usage, sometimes\nmore, so I am not sure whether this is right and hence RFC tag.","testPlan":"","lineCount":"16","dependsOn":[],"reviewers":[],"ccs":["av6","mercurial-devel"],"actions":[{"type":"comment","comment":">>! In D4053#62855, @av6 wrote:\n> Are you sure you're not just using hgwebdir_mod? hgweb_mod is for cases when you don't use --web-conf flag, at least that seems to be the logic that `hgweb.createapp()` uses to pick one of the two modules.\n\nI checked and we are using hgwebdir_mod. When I was testing locally I was not using --web-conf flag, so maybe that's why I ran into `hgweb_mod`. The problem still exists though and we have limited the number of requests at once. Thanks!","author":"pulkit","id":"65973","dateCreated":"1534331436","dateModified":"1534331436"},{"type":"comment","comment":"Are you sure you're not just using hgwebdir_mod? hgweb_mod is for cases when you don't use --web-conf flag, at least that seems to be the logic that `hgweb.createapp()` uses to pick one of the two modules.","author":"av6","id":"62856","dateCreated":"1533231595","dateModified":"1533231595"},{"type":"update","diffId":"9792","author":"pulkit","id":"62851","dateCreated":"1533225478","dateModified":"1533225478"},{"type":"update","diffId":"9787","author":"pulkit","id":"62829","dateCreated":"1533219726","dateModified":"1533219726"}],"dateCreated":"1533219726","dateModified":"1534331436","status":"Needs Review"},{"id":"3639","callsign":"HG","title":"remotenames: add names argument to remotenames revset","author":"pulkit","summary":"This patch adds names argument to the revsets provided by the remotenames\nextension. The revsets are remotenames(), remotebranches() and\nremotebookmarks(). names can be a single names, list of names or can be empty too\nwhich means it's an optional argument.\n\nIf names is\/are passed, changesets which have those remotenames will be\nreturned.\nIf names are not passed, changesets from all the remotenames are shown.\n\nPassing an invalid remotename does not throw error.\n\nThe name argument also supports pattern matching.\n\nTests are added for the argument in tests\/test-logexchange.t","testPlan":"","lineCount":"274","dependsOn":[],"reviewers":[],"ccs":["durin42","yuja","mercurial-devel"],"actions":[{"type":"comment","comment":">>! In D3639#63068, @yuja wrote:\n>> +@revsetpredicate('remotebranches([name, ...])')\n> \n> Can you start with a single \"name\" argument?\n> \n> It doesn't make sense that `remotebranches()` accepts many OR patterns but\n> `branch()` doesn't.\n\nI concur - if you can pass a regex (which you can), that's fine for my use case, let's see how far that gets us.","author":"durin42","id":"63077","dateCreated":"1533306843","dateModified":"1533306843"},{"type":"comment","comment":"> +@revsetpredicate('remotebranches([name, ...])')\n\nCan you start with a single \"name\" argument?\n\nIt doesn't make sense that `remotebranches()` accepts many OR patterns but\n`branch()` doesn't.","author":"yuja","id":"63068","dateCreated":"1533305861","dateModified":"1533305861"},{"type":"update","diffId":"9669","author":"pulkit","id":"61965","dateCreated":"1532810733","dateModified":"1532810733"},{"type":"comment","comment":">>! In D3639#60247, @yuja wrote:\n>>   > > +@revsetpredicate('remotenames([path, ...])')\n>>   >\n>>   > My proposal was `remotenames([pattern])`, just like bookmark(), tag(),\n>>   >  branch(), etc.\n>>   >  If we want a convenient way to specify path prefix, we can\n>>   >  add it to the stringmatcher (e.g. 'remotenames(\"path:server2\")'.)\n>>   \n>>   \n>>   No, I don't want to specify a path prefix. remotenames() will take a path only. Do you want to say that we should allow passing full remotenames instead of just path?\n> \n> Yes, because a remote branch is merely a branch pulled from peer for example,\n> and we'll probably want to select it by -r remotebranch(\"foo\/bar\") just like\n> -r branch(\"foo\/bar\") for normal branches. Do I get it wrong? Why is it so\n> important to filter by a peer path, not by a branch name itself?\n\nYou definitely get it right. I was unable to think that before your last comment because this path approach was motivated from upstream() revsets which I was trying to get rid off. And I like your 'path:' idea too. Thanks!","author":"pulkit","id":"60481","dateCreated":"1530479596","dateModified":"1530479596"},{"type":"comment","comment":">   > > +@revsetpredicate('remotenames([path, ...])')\n>   >\n>   > My proposal was `remotenames([pattern])`, just like bookmark(), tag(),\n>   >  branch(), etc.\n>   >  If we want a convenient way to specify path prefix, we can\n>   >  add it to the stringmatcher (e.g. 'remotenames(\"path:server2\")'.)\n>   \n>   \n>   No, I don't want to specify a path prefix. remotenames() will take a path only. Do you want to say that we should allow passing full remotenames instead of just path?\n\nYes, because a remote branch is merely a branch pulled from peer for example,\nand we'll probably want to select it by -r remotebranch(\"foo\/bar\") just like\n-r branch(\"foo\/bar\") for normal branches. Do I get it wrong? Why is it so\nimportant to filter by a peer path, not by a branch name itself?","author":"yuja","id":"60247","dateCreated":"1530276914","dateModified":"1530276914"},{"type":"comment","comment":">>! In D3639#60108, @yuja wrote:\n>> +@revsetpredicate('remotenames([path, ...])')\n> \n> My proposal was `remotenames([pattern])`, just like bookmark(), tag(),\n> branch(), etc.\n> If we want a convenient way to specify path prefix, we can\n> add it to the stringmatcher (e.g. 'remotenames(\"path:server2\")'.) \n\nNo, I don't want to specify a path prefix. remotenames() will take a path only. Do you want to say that we should allow passing full remotenames instead of just path?\n\n","author":"pulkit","id":"60222","dateCreated":"1530218239","dateModified":"1530218239"},{"type":"comment","comment":"> +@revsetpredicate('remotenames([path, ...])')\n\nMy proposal was `remotenames([pattern])`, just like bookmark(), tag(),\nbranch(), etc. If we want a convenient way to specify path prefix, we can\nadd it to the stringmatcher (e.g. 'remotenames(\"path:server2\")'.) And\nmultiple paths can be specified as 'remotenames(x) + remotenames(y)', though\nit isn't exactly the same as 'remotenames(\"re:<x>|<y>\")'.\n\nDoes it make sense?","author":"yuja","id":"60108","dateCreated":"1530104993","dateModified":"1530104993"},{"type":"update","diffId":"9315","author":"pulkit","id":"60051","dateCreated":"1530049577","dateModified":"1530049577"},{"type":"comment","comment":"I like the literal\/re pattern idea. I think if this lands I can stop maintaining remotebranches, so I'd very much like to see it land.","author":"durin42","id":"57785","dateCreated":"1527794500","dateModified":"1527794500"},{"type":"comment","comment":"> +def _parsepaths(x):\n> +    \"\"\"parses the argument passed in revsets as paths and return\n> +    them as a set or returns None if no path is specified\"\"\"\n> +\n> +    if not x:\n> +        return None\n> +\n> +    paths = set()\n> +    if x[0] in ('symbol', 'string'):\n> +        paths.add(x[1])\n> +    elif x[0] == 'list':\n> +        for t, p in x[1:]:\n> +            paths.add(p)\n> +    return paths\n\nUse `revsetlang.getlist()` and apply `.getstring()` to each element, which\nis more robust.\n\nThat said, the common idiom is to accept a literal\/re pattern. See\n`revset.bookmark()` for example.","author":"yuja","id":"57351","dateCreated":"1527167603","dateModified":"1527167603"},{"type":"update","diffId":"8849","author":"pulkit","id":"57193","dateCreated":"1526851366","dateModified":"1526851366"}],"dateCreated":"1526851366","dateModified":"1533306843","status":"Needs Review"},{"id":"3986","callsign":"HG","title":"split: abort if there's an unfinished operation","author":"martinvonz","summary":"","testPlan":"","lineCount":"10","dependsOn":[],"reviewers":[],"ccs":["quark","mercurial-devel"],"actions":[{"type":"comment","comment":"I think a most flexible solution is to not do the check if there is nothing to rebase.","author":"quark","id":"62080","dateCreated":"1533137720","dateModified":"1533137720"},{"type":"comment","comment":"Oh, I should have said that I agree with not queuing this patch now. We can discuss what we want to do and do it on the default branch later.","author":"martinvonz","id":"61992","dateCreated":"1533018979","dateModified":"1533018979"},{"type":"comment","comment":">>! In D3986#61989, @quark wrote:\n> FB has users reporting they need to split commits in the middle of a histedit. So this might be too restrictive.\n\nTwo thoughts:\n\n1. We had talked at Google about having the \"edit\" action create split markers if the user created multiple commits. Do you think that would address the FB users' use case?\n\n2. It seems very dangerous to not have this check at least when not using obsmarkers and there are descendants of the current commit (split then rebases the descendants), so I'd like to keep the check in at least that case.\n","author":"martinvonz","id":"61991","dateCreated":"1533014407","dateModified":"1533014407"},{"type":"comment","comment":"FB has users reporting they need to split commits in the middle of a histedit. So this might be too restrictive.","author":"quark","id":"61990","dateCreated":"1533014171","dateModified":"1533014171"},{"type":"update","diffId":"9671","author":"martinvonz","id":"61978","dateCreated":"1532987112","dateModified":"1532987112"}],"dateCreated":"1532987112","dateModified":"1533137720","status":"Needs Review"},{"id":"3946","callsign":"HG","title":"obsolete: resolved ValueError for var containing 2 ':' chars(issue-5783)","author":"abhyudaypratap","summary":"feature Note accepts ':' chars\nSo metadata of obsmarker consists of more then 1 ':' char,\nThe content of the variable 'data' in this function is:\n'date:1518013344.878793 -3600\\x00ef1:0\\x00note:6c95ca::a1e17f\\x00p1:ed7673f73387b36521da58a87f08e02e4a795ded\\x00user:Denis Laxalde <xxx@yyy.zz>'\n\nresulting in giving a error ValueError (too many values to unpack) because we are trying to store multiple values into two variables key and value\n\nResolving this error by splitting the variable 'l' from the first occurence of ':'\nkey, value = l.split(\"\", 1)","testPlan":"","lineCount":"2","dependsOn":[],"reviewers":[],"ccs":["yuja","mercurial-devel"],"actions":[{"type":"comment","comment":"Looks good. Can you add a test?\n\nhttps:\/\/www.mercurial-scm.org\/wiki\/ContributingChanges#Coding_style_and_testing\n\nThe version 0 format can be enforced by `--config format.obsstore-version=0`.","author":"yuja","id":"61491","dateCreated":"1531661257","dateModified":"1531661257"},{"type":"update","diffId":"9596","author":"abhyudaypratap","id":"61478","dateCreated":"1531593829","dateModified":"1531593829"}],"dateCreated":"1531593829","dateModified":"1531661257","status":"Needs Review"},{"id":"3830","callsign":"HG","title":"rebase: suppress transaction warns during dry-run","author":"khanchi97","summary":"Suppressed warnings thrown while aborting a transaction during\ndryrun because these warnings does not sound safe to user.","testPlan":"","lineCount":"23","dependsOn":[],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"9593","author":"khanchi97","id":"61465","dateCreated":"1531549446","dateModified":"1531549446"},{"type":"update","diffId":"9501","author":"khanchi97","id":"60911","dateCreated":"1531233875","dateModified":"1531233875"},{"type":"update","diffId":"9384","author":"khanchi97","id":"60399","dateCreated":"1530438740","dateModified":"1530438740"},{"type":"update","diffId":"9359","author":"khanchi97","id":"60261","dateCreated":"1530295767","dateModified":"1530295767"},{"type":"update","diffId":"9347","author":"khanchi97","id":"60219","dateCreated":"1530216657","dateModified":"1530216657"},{"type":"update","diffId":"9269","author":"khanchi97","id":"59837","dateCreated":"1529835539","dateModified":"1529835539"},{"type":"update","diffId":"9267","author":"khanchi97","id":"59827","dateCreated":"1529832650","dateModified":"1529832650"}],"dateCreated":"1529832650","dateModified":"1531549447","status":"Needs Review"},{"id":"3670","callsign":"HG","title":"cmdutil: don't return statedetectionpredicate function from _getrepostate()","author":"pulkit","summary":"This is a part of refactoring which can lead us where we can unify\ncmdutil.STATES and cmdutil.unfinishedstates easily.","testPlan":"","lineCount":"4","dependsOn":["3669"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"plan-changes","author":"pulkit","id":"61257","dateCreated":"1531344976","dateModified":"1531344976"},{"type":"update","diffId":"8925","author":"pulkit","id":"57635","dateCreated":"1527632579","dateModified":"1527632579"}],"dateCreated":"1527632579","dateModified":"1531344976","status":"Changes Planned"},{"id":"3669","callsign":"HG","title":"cmdutil: have statefile names in STATES instead of functions","author":"pulkit","summary":"This will help us in unifying cmdutil.STATES and cmdutil.unfinishedstates as the\nchecking logic for STATES is moved to _getrepostate() and STATES now have just\nfilenames which is similar to unfinishedstates.","testPlan":"","lineCount":"21","dependsOn":[],"reviewers":[],"ccs":["martinvonz","mercurial-devel"],"actions":[{"type":"comment","comment":"I won't have time before the upcoming freeze to get this done.","author":"pulkit","id":"61256","dateCreated":"1531344957","dateModified":"1531344957"},{"type":"plan-changes","author":"pulkit","id":"61255","dateCreated":"1531344957","dateModified":"1531344957"},{"type":"inline","comment":"(Sorry, I missed this one too. I have turned off personal emails from phabricator and use mercurial-devel to keep track of people replying to me.)\n\nThe object sounds good API. I will try that.","replyTo":"58293","isNewFile":"1","line":"3210","lineLength":"0","path":"mercurial\/cmdutil.py","diffId":"8924","author":"pulkit","id":"60483","dateCreated":"1530480466","dateModified":"1530480466"},{"type":"inline","comment":"It's a bit more complicated, I agree, but I think it's better than the alternative. If it removes the need for clearable, it's not that much more complicated.\n\nAn alternative would be to create an object with [is]dirty(), allowcommit(), and clear() methods and have a function that creates such an object for file-based states (all but .hg\/merge?). That would reduce the duplication of fileexistspredicate(<name>) and clearfunction(<name>).\n\nPS. Sorry about the delay. I've been away.","replyTo":"57884","isNewFile":"1","line":"3210","lineLength":"0","path":"mercurial\/cmdutil.py","diffId":"8924","author":"martinvonz","id":"58293","dateCreated":"1528746902","dateModified":"1528746902"},{"type":"inline","comment":"Don't you think it's getting a bit complicated?","replyTo":"57834","isNewFile":"1","line":"3210","lineLength":"0","path":"mercurial\/cmdutil.py","diffId":"8924","author":"pulkit","id":"57884","dateCreated":"1528027688","dateModified":"1528027688"},{"type":"inline","comment":"We don't have to. We could add a `clearstatefn` function in addition to the `statedetectionpredicate`. Perhaps the `if not clearable and ...` stuff could move into the function in that case","replyTo":"57827","isNewFile":"1","line":"3210","lineLength":"0","path":"mercurial\/cmdutil.py","diffId":"8924","author":"martinvonz","id":"57834","dateCreated":"1527872876","dateModified":"1527872876"},{"type":"inline","comment":">How about instead making unfinishedstates more generic by supporting a predicate there\n> instead of the file? We could still allow files as well for BC for a little while.\n\n@martinvonz : because of above util.unlink() we have to keep the filenames in unfinishedstates. :(","replyTo":null,"isNewFile":"1","line":"3210","lineLength":"0","path":"mercurial\/cmdutil.py","diffId":"8924","author":"pulkit","id":"57827","dateCreated":"1527850226","dateModified":"1527850226"},{"type":"inline","comment":"The point of the STATES dict (as opposed to putting all the code here) seems to be to make it generic and extensible. Handling \"merge\" differently here seems like a step backwards. How about instead making unfinishedstates more generic by supporting a predicate there instead of the file? We could still allow files as well for BC for a little while.","replyTo":null,"isNewFile":"1","line":"649","lineLength":"1","path":"mercurial\/cmdutil.py","diffId":"8924","author":"martinvonz","id":"57645","dateCreated":"1527634784","dateModified":"1527634784"},{"type":"update","diffId":"8924","author":"pulkit","id":"57629","dateCreated":"1527632554","dateModified":"1527632554"}],"dateCreated":"1527632554","dateModified":"1531344957","status":"Changes Planned"},{"id":"3672","callsign":"HG","title":"retractboundary: add dryrun parameter","author":"khanchi97","summary":"Added the logic to find those csets whose phase will be changed\nwithout --dry-run and return those csets.","testPlan":"","lineCount":"66","dependsOn":[],"reviewers":[],"ccs":["yuja","pulkit","mercurial-devel"],"actions":[{"type":"inline","comment":"I think affected can be used here. Maybe @yuja might know better.\n\n@yuja we want to return a set of revs whose phase is changed by this function. Does affected looks like the correct set to you?","replyTo":"57830","isNewFile":"1","line":"417","lineLength":"0","path":"mercurial\/phases.py","diffId":"8948","author":"pulkit","id":"61254","dateCreated":"1531344290","dateModified":"1531344290"},{"type":"update","diffId":"8960","author":"khanchi97","id":"57892","dateCreated":"1528034158","dateModified":"1528034158"},{"type":"update","diffId":"8959","author":"khanchi97","id":"57891","dateCreated":"1528033881","dateModified":"1528033881"},{"type":"inline","comment":"okay, will do this for 'advanceboundry' too","replyTo":"57825","isNewFile":"1","line":"511","lineLength":"0","path":"mercurial\/phases.py","diffId":"8948","author":"khanchi97","id":"57833","dateCreated":"1527857439","dateModified":"1527857439"},{"type":"inline","comment":"yeah, I got it.","replyTo":"57824","isNewFile":"1","line":"430","lineLength":"0","path":"mercurial\/phases.py","diffId":"8948","author":"khanchi97","id":"57832","dateCreated":"1527857439","dateModified":"1527857439"},{"type":"inline","comment":"okay","replyTo":"57823","isNewFile":"1","line":"395","lineLength":"0","path":"mercurial\/phases.py","diffId":"8948","author":"khanchi97","id":"57831","dateCreated":"1527857439","dateModified":"1527857439"},{"type":"inline","comment":"Because I think this affected set is calculated after performing actions, when phases are changed. And calculating affected using new phaseroots. Am I missing something here?","replyTo":"57822","isNewFile":"1","line":"417","lineLength":"0","path":"mercurial\/phases.py","diffId":"8948","author":"khanchi97","id":"57830","dateCreated":"1527857439","dateModified":"1527857439"},{"type":"inline","comment":"This should better be:\n\n`If dryrun is True, no actions will be performed\n\nreturns a set of revs whose phase is changed or should be changed`","replyTo":null,"isNewFile":"1","line":"511","lineLength":"0","path":"mercurial\/phases.py","diffId":"8948","author":"pulkit","id":"57825","dateCreated":"1527845546","dateModified":"1527845546"},{"type":"inline","comment":"(Not sure which line I should put this comment on)\n\nJust like D3671, here also, the function should return the correct set of changes whose phase have been changed irrespective of the dryrun value passed.","replyTo":null,"isNewFile":"1","line":"430","lineLength":"0","path":"mercurial\/phases.py","diffId":"8948","author":"pulkit","id":"57824","dateCreated":"1527845546","dateModified":"1527845546"},{"type":"inline","comment":"Add documentation about dry-run and the return value.","replyTo":null,"isNewFile":"1","line":"395","lineLength":"0","path":"mercurial\/phases.py","diffId":"8948","author":"pulkit","id":"57823","dateCreated":"1527845546","dateModified":"1527845546"},{"type":"inline","comment":"Why are we not using this affected set here to find the changesets whose phase is changed?","replyTo":null,"isNewFile":"1","line":"417","lineLength":"0","path":"mercurial\/phases.py","diffId":"8948","author":"pulkit","id":"57822","dateCreated":"1527845546","dateModified":"1527845546"},{"type":"update","diffId":"8948","author":"khanchi97","id":"57812","dateCreated":"1527836390","dateModified":"1527836390"},{"type":"update","diffId":"8929","author":"khanchi97","id":"57663","dateCreated":"1527675912","dateModified":"1527675912"}],"dateCreated":"1527675912","dateModified":"1531344290","status":"Needs Review"},{"id":"3866","callsign":"HG","title":"histedit: use cbor to write histedit-state file","author":"pulkit","summary":"This patch starts using state.cmdstate.write() which uses cbor as serialization\nformat to write histedit-state files.\n\nCompatibility for reading old state files using new code is maintained.\n\nSome changes in test which depended on the format of histedit state file.","testPlan":"","lineCount":"64","dependsOn":[],"reviewers":["durin42"],"ccs":["yuja","mercurial-devel"],"actions":[{"type":"comment","comment":"I am not confident to land some state format changes at this point in the cycle. Let's do it early in next cycle.","author":"pulkit","id":"60839","dateCreated":"1531161109","dateModified":"1531161109"},{"type":"plan-changes","author":"pulkit","id":"60838","dateCreated":"1531161109","dateModified":"1531161109"},{"type":"comment","comment":">>! In D3866#60421, @yuja wrote:\n> Queued the first 4 patches, thanks.\n> \n>>    $ HGEDITOR=\"sh $TESTTMP\/editplan.sh\" hg histedit --edit-plan\n>>    $ cat .hg\/histedit-state\n>> -  v1\n>> -  055a42cdd88768532f9cf79daa407fc8d138de9b\n>> -  3c6a8ed2ebe862cc949d2caa30775dd6f16fb799\n>> -  False\n>> -  3\n>> -  drop\n>> -  e860deea161a2f77de56603b340ebbb4536308ae\n>> -  drop\n>> -  652413bf663ef2a641cab26574e46d5f5a64a55a\n>> -  drop\n>> -  3c6a8ed2ebe862cc949d2caa30775dd6f16fb799\n>> -  0\n>> -  strip-backup\/177f92b77385-0ebe6a8f-histedit.hg\n>> +  2\n> \n> I know anything other than \"v1\" will make old hg crash, but I think the\n> version syntax should stay \"v%d\" (i.e. \"v2\" in this case.)\n\nYeah, I think we need to pass version number in the state.cmdstate.read(). Supporting rebasestate will also need similar tweak.\n> \n>>      def _write(self, fp):\n>> -        fp.write('v1\\n')\n>> -        fp.write('%s\\n' % node.hex(self.parentctxnode))\n>> -        fp.write('%s\\n' % node.hex(self.topmost))\n>> -        fp.write('%s\\n' % ('True' if self.keep else 'False'))\n>> -        fp.write('%d\\n' % len(self.actions))\n>> +        data = {'parentctxnode': self.parentctxnode,\n>> +                'topmost': self.topmost,\n>> +                'keep': self.keep,\n>> +                'backupfile': self.backupfile,\n>> +                'replacements': self.replacements}\n>> +        rules = []\n>>          for action in self.actions:\n>> -            fp.write('%s\\n' % action.tostate())\n>> -        fp.write('%d\\n' % len(self.replacements))\n>> -        for replacement in self.replacements:\n>> -            fp.write('%s%s\\n' % (node.hex(replacement[0]), ''.join(node.hex(r)\n>> -                for r in replacement[1])))\n>> -        backupfile = self.backupfile\n>> -        if not backupfile:\n>> -            backupfile = ''\n>> -        fp.write('%s\\n' % backupfile)\n>> +            rules.append(action.tostate().split('\\n'))\n>> +        rules = \"\\n\".join([\"%s %s\" % (verb, rest) for [verb, rest] in rules])\n> \n> Isn't it better to serialize `[(verb, rest)]` as is since CBOR can store that?\n\nYeah, it's better. I will do that.","author":"pulkit","id":"60480","dateCreated":"1530479460","dateModified":"1530479460"},{"type":"comment","comment":"Queued the first 4 patches, thanks.\n\n>    $ HGEDITOR=\"sh $TESTTMP\/editplan.sh\" hg histedit --edit-plan\n>    $ cat .hg\/histedit-state\n> -  v1\n> -  055a42cdd88768532f9cf79daa407fc8d138de9b\n> -  3c6a8ed2ebe862cc949d2caa30775dd6f16fb799\n> -  False\n> -  3\n> -  drop\n> -  e860deea161a2f77de56603b340ebbb4536308ae\n> -  drop\n> -  652413bf663ef2a641cab26574e46d5f5a64a55a\n> -  drop\n> -  3c6a8ed2ebe862cc949d2caa30775dd6f16fb799\n> -  0\n> -  strip-backup\/177f92b77385-0ebe6a8f-histedit.hg\n> +  2\n\nI know anything other than \"v1\" will make old hg crash, but I think the\nversion syntax should stay \"v%d\" (i.e. \"v2\" in this case.)\n\n>      def _write(self, fp):\n> -        fp.write('v1\\n')\n> -        fp.write('%s\\n' % node.hex(self.parentctxnode))\n> -        fp.write('%s\\n' % node.hex(self.topmost))\n> -        fp.write('%s\\n' % ('True' if self.keep else 'False'))\n> -        fp.write('%d\\n' % len(self.actions))\n> +        data = {'parentctxnode': self.parentctxnode,\n> +                'topmost': self.topmost,\n> +                'keep': self.keep,\n> +                'backupfile': self.backupfile,\n> +                'replacements': self.replacements}\n> +        rules = []\n>          for action in self.actions:\n> -            fp.write('%s\\n' % action.tostate())\n> -        fp.write('%d\\n' % len(self.replacements))\n> -        for replacement in self.replacements:\n> -            fp.write('%s%s\\n' % (node.hex(replacement[0]), ''.join(node.hex(r)\n> -                for r in replacement[1])))\n> -        backupfile = self.backupfile\n> -        if not backupfile:\n> -            backupfile = ''\n> -        fp.write('%s\\n' % backupfile)\n> +            rules.append(action.tostate().split('\\n'))\n> +        rules = \"\\n\".join([\"%s %s\" % (verb, rest) for [verb, rest] in rules])\n\nIsn't it better to serialize `[(verb, rest)]` as is since CBOR can store that?","author":"yuja","id":"60422","dateCreated":"1530448552","dateModified":"1530448552"},{"type":"update","diffId":"9378","author":"pulkit","id":"60359","dateCreated":"1530371419","dateModified":"1530371419"}],"dateCreated":"1530371419","dateModified":"1531161109","status":"Changes Planned"},{"id":"3771","callsign":"HG","title":"py3: make contrib\/perf.py compatible to Python 3","author":"pulkit","summary":"This patch makes contrib\/perf.py compatible on Python 3 by using\npycompat.byteskwargs(), pycompat.sysstr(), replacing sys.maxint with sys.maxsize\nbecause sys.maxint is not present on Python 3, adding conditional for\nxrange\/range and adding a lot of b'' prefixes.\n\nThis makes test-contrib-perf.t passes on Python 3.","testPlan":"","lineCount":"481","dependsOn":[],"reviewers":[],"ccs":["yuja","mercurial-devel"],"actions":[{"type":"comment","comment":"> +    fm.write(b'wall', ' wall %f', m[0])\n> +    fm.write(b'comb', ' comb %f', m[1] + m[2])\n> +    fm.write(b'user', ' user %f', m[1])\n> +    fm.write(b'sys',  ' sys %f', m[2])\n> +    fm.write(b'count',  ' (best of %d)', count)\n\nNeeds more b''s.\n\nI think it's better to split patches to b'' and the other. I'll send the\npatch series that extracts the b'' transformer from the module importer.","author":"yuja","id":"59320","dateCreated":"1529327086","dateModified":"1529327086"},{"type":"update","diffId":"9140","author":"pulkit","id":"59083","dateCreated":"1529263218","dateModified":"1529263218"}],"dateCreated":"1529263218","dateModified":"1529327086","status":"Needs Review"},{"id":"3650","callsign":"HG","title":"serve: add an option to open in the default browser","author":"nspanti-logilab","summary":"It is inspired by a similar functionality in pydoc. `-g`\/`--graphical` may not be adequate, the only other idea that I have is `-w`\/`--window`.","testPlan":"","lineCount":"42","dependsOn":[],"reviewers":[],"ccs":["yuja","pulkit","av6","durin42","mercurial-devel"],"actions":[{"type":"comment","comment":"(Resend because I forgot to remove the \"someone wrote:\" line again, sorry.)\n\n>   For laziness,  this patch could be also useful. It is also better for text shell beginners, because they may not know that they can combine commands like that (even if it is basic).\n\nFor laziness, we can add it to `[alias]`. I'm not a fan of adding lots of\none-off options since that's the source of man page bloat.\n\nHow about document the xdg-open example in `hg help serve -v`?","author":"yuja","id":"58867","dateCreated":"1529155285","dateModified":"1529155285"},{"type":"comment","comment":"For laziness,  this patch could be also useful. It is also better for text shell beginners, because they may not know that they can combine commands like that (even if it is basic). Moreover, you can still use `hg serve` in foreground or use it in background and give the result of `--print-url` to a non web browser application (but I imagine that there is a way to duplicate stdout to multiple programs). These are the advantages of this patch that I see over `--print-url` of D3649. It is better for convenience and require less knowledge for the user, but it is not a need. D3649 was precisely suggested to me because it can fulfill the functionality of the current patch and is more versatile.","author":"nspanti-logilab","id":"58735","dateCreated":"1529052032","dateModified":"1529052032"},{"type":"comment","comment":"Do we need this patch any more since D3649 has landed and we can do `xdg-open \"$(hg serve --print-url)\"`?","author":"pulkit","id":"58567","dateCreated":"1529005653","dateModified":"1529005653"},{"type":"comment","comment":"D3649 mentions `xdg-open \"$(hg serve --print-url)\"`, which shows that --print-url can nearly replace --graphical and is more versatile.","author":"av6","id":"57842","dateCreated":"1527944680","dateModified":"1527944680"},{"type":"comment","comment":"I don't love the flag name, and am -0 on the feature in general. All terminal emulators I've used in the last 10 years have URL detection logic and make clicking the link trivial...","author":"durin42","id":"57770","dateCreated":"1527794132","dateModified":"1527794132"},{"type":"update","diffId":"8881","author":"nspanti-logilab","id":"57359","dateCreated":"1527168904","dateModified":"1527168904"}],"dateCreated":"1527168904","dateModified":"1529155285","status":"Needs Review"},{"id":"2945","callsign":"HG","title":"state: add a magicheader to each state file","author":"pulkit","summary":"This patch adds a magic header to each state file written using CBOR format so\nthat we can distinguish between old state files and new one even if old state\nfiles get successfully parsed by cbor.","testPlan":"","lineCount":"1","dependsOn":[],"reviewers":[],"ccs":["jeffpc","yuja","mercurial-devel"],"actions":[{"type":"comment","comment":">>! In D2945#58586, @pulkit wrote:\n> There is a suggestion from @jeffpc to have a plain text header in statefiles like `HGStateFile` so that user can grep for them or look into files and know that they are state files. How does that sound?\n> (@jeffpc I am not sure I explained this correctly, it will be great if you can chime in.)\n\nI don't really care about the actual string value, I just think it is a good storage design practice to have every on-disk structure contain a magic value of some sort.  (Beware, I have a storage background and therefore opinions about these things... ;) )  This makes it easier to do all sort of data recovery should the need arise.  E.g.,\n\n  - if you know that a random file (e.g., from lost+found) is a hg state file, you can ignore\/delete it and just reclone your repos\n  - hg can detect some corruption better, and depending on the state file regenerate it or error out\n  - hg developers can more easily sort out what's what when looking at buffers\/files\/etc.\n  - as an added bonus, the magic can be used for format versioning\n\n\nFWIW, the three major things that I consider essential to good on-disk structures:\n\n  # magic numbers - see above\n  # checksums - possibly two checksums: one for the header and one for the payload\n  # back-pointers - see below\n\nBack-pointers allow you to tie the file\/buffer into the bigger context - be it which state file it is, or which repository it belongs to.  In this case a state file's back-pointer could be something like the repo-id + state-file-name stored in the state file's header.  (I realize that there is no repo-id, but in general some approximation of it tends to be better than nothing.)  This should make it painfully obvious what data one is dealing with and how to parse\/interpret it just by looking at the byte-stream without any other context.\n\n\nBy the way, I think that the magic should *not* be part of the CBOR payload but rather be at a fixed location in the file.  The easier it is to find, the easier it is to detect issues (and to avoid trying to decode bad\/malicious input).","author":"jeffpc","id":"58725","dateCreated":"1529024379","dateModified":"1529024509"},{"type":"comment","comment":"There is a suggestion from @jeffpc to have a plain text header in statefiles like `HGStateFile` so that user can grep for them or look into files and know that they are state files. How does that sound?\n(@jeffpc I am not sure I explained this correctly, it will be great if you can chime in.)","author":"pulkit","id":"58587","dateCreated":"1529007034","dateModified":"1529007034"},{"type":"comment","comment":"> There are some old state files, which don't have a version header on top\n> of them like graftstate. I need suggestion on how to handle them.\n\nIn that case, we would have to either do some heuristic (like histedit), or maintain\nold\/new state files (like mergestate.) Perhaps graft can do the latter?","author":"yuja","id":"47899","dateCreated":"1522244380","dateModified":"1522244380"},{"type":"comment","comment":">>! In D2945#47831, @yuja wrote:\n> I think the magic has to vary depending on the current state file format.\n> The state file must be backward\/forward compatible with the older\/newer formats.\n> \n> If the current magic is `2\\n` (and is parsed as `int(f.readline())`) for example, the\n> CBOR preamble would have to be `3\\n`. Otherwise, old hg client cant determine\n> whether the state file is corrupted or written by newer client.\n\nThere are some old state files, which don't have a version header on top of them like graftstate. I need suggestion on how to handle them.\n","author":"pulkit","id":"47892","dateCreated":"1522235138","dateModified":"1522235138"},{"type":"comment","comment":"I think the magic has to vary depending on the current state file format.\nThe state file must be backward\/forward compatible with the older\/newer formats.\n\nIf the current magic is `2\\n` (and is parsed as `int(f.readline())`) for example, the\nCBOR preamble would have to be `3\\n`. Otherwise, old hg client cant determine\nwhether the state file is corrupted or written by newer client.\n\nFWIW, last time I reviewed `simplekeyvaluefile`, I insisted that a parser should\ntake a file stream instead of a filename, so the caller can handle compatibility thingy.\n\n```\nwith open(\"statefile\", \"rb\") as fp:\n    try:\n        version = int(fp.readline())\n    except ValueError:\n        raise some nicer exception\n    if version == 2:\n        return readold(fp)\n    elif version == 3:\n        return readcbor(fp)\n```","author":"yuja","id":"47832","dateCreated":"1522155487","dateModified":"1522155487"},{"type":"update","diffId":"7299","author":"pulkit","id":"47624","dateCreated":"1522084159","dateModified":"1522084159"}],"dateCreated":"1522084159","dateModified":"1529024379","status":"Needs Review"},{"id":"3684","callsign":"HG","title":"shelve: merge in obsshelve changes implemented at facebook","author":"durin42","summary":"test-obsshelve.t could be merged with test-shelve.t at some point, but\nfor now I want the comprehensive coverage. Note that obsshelve in\nhg-experimental is a fork of shelve, so we're just merging things back\ntogether. The path to graduation from experimental for obsshelve is to\nhave a better solution around hiding revisions without creating a ton\nof markers that'll get exchanged. We've spent a fair amount of time\ntalking about what that should look like, but in the interim the\nexperimental-branded obsshelve is a good thing we should have in core.\n\nAs a follow-up I'd like to make obsshelve a case of test-shelve.t, but\nI thought this would be easier for the initial import.","testPlan":"","lineCount":"1759","dependsOn":[],"reviewers":[],"ccs":["smf","lothiraldan","mercurial-devel"],"actions":[{"type":"comment","comment":"durin42 (Augie Fackler) <phabricator@mercurial-scm.org> writes:\n\n> durin42 created this revision.\n> Herald added a subscriber: mercurial-devel.\n> Herald added a reviewer: hg-reviewers.\n>\n> REVISION SUMMARY\n>   test-obsshelve.t could be merged with test-shelve.t at some point, but\n>   for now I want the comprehensive coverage. Note that obsshelve in\n>   hg-experimental is a fork of shelve, so we're just merging things back\n>   together. The path to graduation from experimental for obsshelve is to\n>   have a better solution around hiding revisions without creating a ton\n>   of markers that'll get exchanged. We've spent a fair amount of time\n>   talking about what that should look like, but in the interim the\n>   experimental-branded obsshelve is a good thing we should have in core.\n>\n>   As a follow-up I'd like to make obsshelve a case of test-shelve.t, but\n>   I thought this would be easier for the initial import.\n\nI'm trying to catch up on this discussion, so apologies if this is the\nwrong thread to reply to.\n\nI've been mulling on this problem for quite some time now. To clarify,\nI'm referring to the problem of hiding \/ stripping changesets that are\nknown not to be shared. I keep coming back to the idea of a new phase\nbecause it is so elegant. Looking at the other patch series, I am sad we\ndidn't implement the phases using the < operator and instead testing\ndirectly for the name of the phase.\n\nHaving a \"internal-only\" phase makes a lot of sense for temporary\ncommits that only hg is making. In fact, I'd like to suggest taking it\neven further: not making obs markers for anything above the draft phase\n(i.e. secret, internal, etc.). This would push development of wip\nbranches to perhaps be secret instead of the default draft.\n\nGlancing at the other discussions in IRC and on the list, it seems we're\nmostly agreed on the new phase?\n\nP.S. I think I adjusted my email filters to put phabricator emails I'm\ntagged in directly into my inbox; so replying to this message would help\ntest that (even if just to say, \"we should this discussion to another\nthread\")\n\n  - {F102400, layout=link}","author":"smf","id":"58618","dateCreated":"1529017416","dateModified":"1529017416"},{"type":"comment","comment":">>! In D3684#58106, @durin42 wrote:\n>>>! In D3684#58074, @lothiraldan wrote:\n>> We have an implementation of shelve without strip nor obsmarkers.\n> \n> Can we see that now? Otherwise I'm going to lobby to land this in the name of having something done.\n\nI pushed the full series as RFC on the mailing list (https:\/\/www.mercurial-scm.org\/pipermail\/mercurial-devel\/2018-June\/117682.html) as I was not sure how phabricator would handle the fact that there is two independent parts + one part that needs the two other. I sent it as linear to more easily see the big picture but the first part of internal phase can be merged independently.\n\nI plan to send the next part of the series after the first part has landed for flow control.","author":"lothiraldan","id":"58112","dateCreated":"1528381092","dateModified":"1528381092"},{"type":"comment","comment":">>! In D3684#58074, @lothiraldan wrote:\n> We have an implementation of shelve without strip nor obsmarkers.\n\nCan we see that now? Otherwise I'm going to lobby to land this in the name of having something done.","author":"durin42","id":"58106","dateCreated":"1528315038","dateModified":"1528315038"},{"type":"comment","comment":"After reading your obsshelve series, it seems that we have some overlap.\n\nWe have an implementation of shelve without strip nor obsmarkers.\n\nThe first series https:\/\/phab.mercurial-scm.org\/D3685 is the first half, cleaning up the shelve extensions in order to simplify the code flow and gain access to more data.\n\nThe second half implements the strip less version. We wanted to wait until the first half has been merged but we can send the whole series if that can help the discussion. The proposed implementation is based on the internal phase plan.\n\nThe cleaning refactoring series seems worthy on its own but will results in conflicts. Do you think it would be possible to rebase your obsshelve serie on top of our refactoring? This way we could have a clean obsshelve code and you should be able to remove some of the hacks in your obsshelve serie.\n\n","author":"lothiraldan","id":"58074","dateCreated":"1528226505","dateModified":"1528226505"},{"type":"comment","comment":"We just sent https:\/\/phab.mercurial-scm.org\/D3685 that update shelve to not use rebase anymore. ","author":"lothiraldan","id":"58063","dateCreated":"1528196391","dateModified":"1528196391"},{"type":"update","diffId":"8970","author":"durin42","id":"57953","dateCreated":"1528150570","dateModified":"1528150570"}],"dateCreated":"1528150570","dateModified":"1529017416","status":"Needs Review"},{"id":"1313","callsign":"HG","title":"dirstate: gate access to self._map.dirs","author":"mbolin","summary":"`_droppath()` already checks `\"dirs\" in self._map.__dict__` before accessing\n`self._map.dirs`. This updates `_addpath()` to be equally cautious.","testPlan":"","lineCount":"23","dependsOn":[],"reviewers":["mbthomas"],"ccs":["pulkit","mbthomas","durham","mercurial-devel"],"actions":[{"type":"comment","comment":"Do we need this anymore or can this be abandoned? (trying to close some not-required-any-more differentials)","author":"pulkit","id":"58571","dateCreated":"1529006100","dateModified":"1529006100"},{"type":"inline","comment":"I don't think it's safe to do this.  We're looking for clashes between files and directories, so we do need to look in self.dirs, even if that means generating it because we're accessing it for the first time.","replyTo":null,"isNewFile":"1","line":"405","lineLength":"0","path":"mercurial\/dirstate.py","diffId":"3267","author":"mbthomas","id":"21733","dateCreated":"1509750955","dateModified":"1509750955"},{"type":"reject","author":"mbthomas","id":"21732","dateCreated":"1509750955","dateModified":"1509750955"},{"type":"update","diffId":"3267","author":"mbolin","id":"21716","dateCreated":"1509749978","dateModified":"1509749978"}],"dateCreated":"1509749978","dateModified":"1529006100","status":"Needs Revision"},{"id":"3664","callsign":"HG","title":"graft: reuse the --edit value passed initially in `hg graft --continue` (BC)","author":"pulkit","summary":"This patch starts storing the value of --edit flag passed by the user initially\nwhen they runs `hg graft --edit`. We can now read the value from graftstate in\ncase of `--continue` and respect the value which user passed earlier.\n\nThe tests changes demonstrate the fix.","testPlan":"","lineCount":"28","dependsOn":["3663"],"reviewers":[],"ccs":["yuja","mercurial-devel"],"actions":[{"type":"comment","comment":"Queued the first two, thanks.\n\n>   This patch starts storing the value of --edit flag passed by the user initially\n>   when they runs `hg graft --edit`. We can now read the value from graftstate in\n>   case of `--continue` and respect the value which user passed earlier.\n\nI'm -0.5 on this since --edit is an option to change the interactive behavior,\nand I think we don't preserve such options across interrupted sessions.","author":"yuja","id":"58065","dateCreated":"1528200002","dateModified":"1528200002"},{"type":"update","diffId":"8909","author":"pulkit","id":"57568","dateCreated":"1527287139","dateModified":"1527287139"}],"dateCreated":"1527287139","dateModified":"1528200002","status":"Needs Review"},{"id":"3187","callsign":"HG","title":"phase: add dry-run functionality","author":"khanchi97","summary":"Added the logic to find those csets whose phase will be changed\n(while running without --dry-run). And show the list of those csets.","testPlan":"","lineCount":"173","dependsOn":[],"reviewers":["pulkit"],"ccs":["av6","mercurial-devel"],"actions":[{"type":"update","diffId":"8962","author":"khanchi97","id":"57894","dateCreated":"1528037274","dateModified":"1528037274"},{"type":"comment","comment":"Nice work! I like where how you simplified things from previous version.\n\nIf you look at the patch, this seems to do multiple things at once which are:\n\n  * adding dry-run argument to advanceboundary and retractboundary functions\n  * adding a new --dry-run flag to hg phase\n\nLet's break things up in individual patches so that we can improve more on individual pieces. Can you split this into a separate patch for adding dryrun argument to each function, i.e. one patch for adcanceboundary, one for retractboundary, and then one last patch which adds the new flag? That will be very much helpful in reviewing this work and also making incremental changes and moving forward step by step.","author":"pulkit","id":"57621","dateCreated":"1527622549","dateModified":"1527622549"},{"type":"comment","comment":"@pulkit Now I moved the logic for finding revs (those phase will change)  to advanceboundry and retractboundry function and for now it shows all nodes. I am working to show range instead. Take a look at this when you are free :)","author":"khanchi97","id":"57619","dateCreated":"1527584154","dateModified":"1527584154"},{"type":"update","diffId":"8896","author":"khanchi97","id":"57476","dateCreated":"1527280427","dateModified":"1527280427"},{"type":"comment","comment":">>! In D3187#57349, @khanchi97 wrote:\n> okay, but I have some queries like\n> 1. How about showing revision no. instead of cset id?\n\nI think we should show cset ids. If you want to with rev numbers, go with that. This should not be a blocker to get the initial patch in.\n> 2. And in this https:\/\/pastebin.com\/raw\/kWcr9xVK example if I change revision 2 phase to --secret then how it should print the range, I mean now we have branches in this range?\n> \n> Can we show that range like this:\n> \n> 0316ce92851d : : b385d13d5ed4  draft  ->  secret\n> 0316ce92851d : : 4ccc844d5454   draft  -> secret\n\nThere are two ways you can show the range in this case:\n\n1) 0316ce92851d::b385d13d5ed4 and f19b7f89f44e::4ccc844d5454\n2) 0316ce92851d::4ccc844d5454 and b385d13d5ed4\n\nLook into `hg help revsets` to understand what `::` means.","author":"pulkit","id":"57419","dateCreated":"1527196739","dateModified":"1527196739"},{"type":"comment","comment":"okay, but I have some queries like\n1. How about showing revision no. instead of cset id?\n2. And in this https:\/\/pastebin.com\/raw\/kWcr9xVK example if I change revision 2 phase to --secret then how it should print the range, I mean now we have branches in this range?\n\nCan we show that range like this:\n\n0316ce92851d : : b385d13d5ed4  draft  ->  secret\n0316ce92851d : : 4ccc844d5454   draft  -> secret","author":"khanchi97","id":"57349","dateCreated":"1527158527","dateModified":"1527158527"},{"type":"inline","comment":"This is not very much helpful. In such cases, how about showing the range of changesets, something like:\n```f7b1eb17ad24::925d80f479bb public -> draft\nb385d13d5ed4::fdc0253c25cf secret -> draft``` ","replyTo":null,"isNewFile":"1","line":"990","lineLength":"0","path":"tests\/test-phases.t","diffId":"7877","author":"pulkit","id":"57180","dateCreated":"1526844970","dateModified":"1526844970"},{"type":"inline","comment":"Since all the logic is copy-pasted from phases.advanceboundary and phases.retractboundary, why not we just pass the `--dry-run` option there and get the count from those functions only.","replyTo":null,"isNewFile":"1","line":"3898","lineLength":"0","path":"mercurial\/commands.py","diffId":"7877","author":"pulkit","id":"57179","dateCreated":"1526844970","dateModified":"1526844970"},{"type":"comment","comment":"@pulkit Can you please review it?","author":"khanchi97","id":"51519","dateCreated":"1523353338","dateModified":"1523353338"},{"type":"update","diffId":"7877","author":"khanchi97","id":"51174","dateCreated":"1523185235","dateModified":"1523185235"},{"type":"inline","comment":"`\"cannot use --dry-run without target phase\"`","replyTo":null,"isNewFile":"1","line":"3872","lineLength":"0","path":"mercurial\/commands.py","diffId":"7874","author":"av6","id":"51173","dateCreated":"1523167056","dateModified":"1523167056"},{"type":"update","diffId":"7874","author":"khanchi97","id":"51159","dateCreated":"1523101349","dateModified":"1523101349"},{"type":"update","diffId":"7873","author":"khanchi97","id":"51153","dateCreated":"1523100805","dateModified":"1523100805"}],"dateCreated":"1523100805","dateModified":"1528037274","status":"Needs Review"},{"id":"3445","callsign":"HG","title":"dispatch: validate return type from dispatch() return value","author":"indygreg","summary":"This commit adds type checking to the result of dispatch() so\nwe catch non-None and non-int values. This prevents crashes\nwhen we later try to \"& 255\" this value.\n\nThe Python 3 test failures introduced previously have all gone away.","testPlan":"","lineCount":"50","dependsOn":["3444"],"reviewers":["admin","durin42"],"ccs":["durin42","admin","yuja","mercurial-devel"],"actions":[{"type":"comment","comment":"(derp, that was me, but my password vault mis-fired and I got admin instead of me.)","author":"durin42","id":"57766","dateCreated":"1527793864","dateModified":"1527793864"},{"type":"reject","author":"durin42","id":"57765","dateCreated":"1527793864","dateModified":"1527793864"},{"type":"comment","comment":"This looks good, but needs rebased.","author":"admin","id":"57762","dateCreated":"1527793807","dateModified":"1527793807"},{"type":"reject","author":"admin","id":"57761","dateCreated":"1527793807","dateModified":"1527793807"},{"type":"update","diffId":"8654","author":"indygreg","id":"56035","dateCreated":"1526100492","dateModified":"1526100492"},{"type":"comment","comment":"IIRC, SystemExit is caught at callcatch and translated to a status code.\n\n> +    # The logic here attempts to mimic how CPython's pythonrun.c does things.\n> +    # Essentially:\n> +    #\n> +    # * Exit is controlled by returning a value or raising SystemExit.\n> +    # * An integer value exits with that exit code.\n> +    # * A value of ``None`` is interpreted as ``0``.\n> +    # * A non-integer, non-None value results in that value being printed\n> +    #   and an exit code of ``1``.\n> +    # * SystemExit can have a ``code`` attribute containing the exit value.\n>      try:\n> -        status = (dispatch(req) or 0)\n> +        code = dispatch(req) or 0\n>      except error.StdioError as e:\n>          err = e\n> -        status = -1\n> +        code = -1\n> +    except SystemExit as err:\n> +        try:\n> +            code = err.code\n> +            err = None\n> +        except AttributeError:","author":"yuja","id":"54925","dateCreated":"1525609762","dateModified":"1525609762"},{"type":"update","diffId":"8460","author":"indygreg","id":"54814","dateCreated":"1525579872","dateModified":"1525579872"}],"dateCreated":"1525579872","dateModified":"1527794332","status":"Needs Revision"},{"id":"3444","callsign":"HG","title":"tests: comprehensively test exit handling","author":"indygreg","summary":"Rust `hg` currently has some test failures around exit handling.\nThis commit establishes some centralized tests around exit handling so\nthat we can unify the behavior of our Rust frontend and `python`.\n\nThere are some added tests that use a value other than an integer\nor None for the exit code. The docs for sys.exit() say such a value\nis allowed. However, Mercurial currently crashes in this case. Upcoming\ncommits will teach Mercurial to handle these values.\n\nThis does introduce a few Python 3 test failures. However, this test\nalready has a few failures. And the failures being introduced should\nmostly go away with subsequent commits. So I think this is acceptable.","testPlan":"","lineCount":"91","dependsOn":[],"reviewers":["durin42"],"ccs":["durin42","yuja","mercurial-devel"],"actions":[{"type":"comment","comment":"This LG, esp with D3445 which appears to be its child, but D3445 still needs rebased AFAICT","author":"durin42","id":"57781","dateCreated":"1527794332","dateModified":"1527794332"},{"type":"accept","author":"durin42","id":"57779","dateCreated":"1527794332","dateModified":"1527794332"},{"type":"comment","comment":">   > I generally like the direction of this series, but I think there's no point\n>   >  to extend Mercurial's exit code handling to support all weird Python types.\n>   >\n>   > Only ints and (None for 0) are ever valid.\n>   \n>   \n>   The reason I did this is because from the context or `rhg`, we don't have CPython's default exit handling to fall back on. Our choices are:\n>   \n>   1. Reimplement CPython's exit\/error handling in Rust\n>   2. Reimplement CPython's exit\/error handling in Python in dispatch [so the end state is more well-defined]\n>   3. Do something crude in `rhg` when we hit special cases that CPython would normally deal with.\n>   \n>   Since extensions could do weird things, I figured 2 was the best option.\n\nI don't fully understand the situation in `rhg`, but maybe we can instead\nfix `scmutil.callcatch()` to either translate non-int SystemExit to integer\nor raise ProgrammingError as a bad use of sys.exit() in hg. No SystemExit\nshould be raised outside of the callcatch().\n\nI'm against allowing command functions to return any objects other than\nint or None because it's useless.","author":"yuja","id":"56045","dateCreated":"1526104337","dateModified":"1526104337"},{"type":"comment","comment":">>! In D3444#54922, @yuja wrote:\n> I generally like the direction of this series, but I think there's no point\n> to extend Mercurial's exit code handling to support all weird Python types.\n> \n> Only ints and (None for 0) are ever valid.\n\nThe reason I did this is because from the context or `rhg`, we don't have CPython's default exit handling to fall back on. Our choices are:\n\n1. Reimplement CPython's exit\/error handling in Rust\n2. Reimplement CPython's exit\/error handling in Python in dispatch [so the end state is more well-defined]\n3. Do something crude in `rhg` when we hit special cases that CPython would normally deal with.\n\nSince extensions could do weird things, I figured 2 was the best option.\n\nBut I'll take a look at this series again and re-evaluate if we can simplify things. I just thought I'd brain dump on what I'm thinking.","author":"indygreg","id":"56014","dateCreated":"1526097317","dateModified":"1526097317"},{"type":"comment","comment":"I generally like the direction of this series, but I think there's no point\nto extend Mercurial's exit code handling to support all weird Python types.\n\nOnly ints and (None for 0) are ever valid.","author":"yuja","id":"54923","dateCreated":"1525609756","dateModified":"1525609756"},{"type":"update","diffId":"8459","author":"indygreg","id":"54808","dateCreated":"1525579871","dateModified":"1525579871"}],"dateCreated":"1525579871","dateModified":"1527794332","status":"Accepted"},{"id":"3454","callsign":"HG","title":"Assign to result from block.","author":"kevincox","summary":"This is more idiomatic because you don't have uninitialized variables and has less typing.","testPlan":"","lineCount":"9","dependsOn":["3447"],"reviewers":["indygreg","durin42"],"ccs":["martinvonz","durin42","mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"57759","dateCreated":"1527793653","dateModified":"1527793653"},{"type":"comment","comment":"It looks like this is based on top of D3447. So I'll just hook up the dependency and we'll land it as soon as my series is landed.","author":"indygreg","id":"56011","dateCreated":"1526097062","dateModified":"1526097062"},{"type":"accept","author":"indygreg","id":"56009","dateCreated":"1526097062","dateModified":"1526097062"},{"type":"comment","comment":"This doesn't apply. Looks like it's based on a version where exit code is 1 (line 193), but our main.rs seems to have always have had it set to 255.","author":"martinvonz","id":"55724","dateCreated":"1525900608","dateModified":"1525900608"},{"type":"update","diffId":"8471","author":"kevincox","id":"54912","dateCreated":"1525594657","dateModified":"1525594657"}],"dateCreated":"1525594657","dateModified":"1527793653","status":"Accepted"},{"id":"3447","callsign":"HG","title":"rust: make exit handling consistent with `hg`","author":"indygreg","summary":"Now that dispatch.run() handles special exit cases and always returns\nan exit code, Rust's (formerly lacking) invocation of dispatch.run() can\nhandle the return value with minimal hassle.\n\nIn addition, we change Rust to exit 1 instead of 255 in the case\nof unhandled errors, as that is actually what Python does.\n\nThis fixes a few test failures when running the test suite with rhg.","testPlan":"","lineCount":"22","dependsOn":["3446"],"reviewers":["kevincox","durin42"],"ccs":["durin42","kevincox","mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"57757","dateCreated":"1527793647","dateModified":"1527793647"},{"type":"inline","comment":"s\/unreachable\/panic\/ since the code can never be hit as long as the python code returns the right type.\n\nAlso it might be a good idea to try to print out the value returned in the error message. Although if not easy it is probably not worth it.","replyTo":null,"isNewFile":"1","line":"204","lineLength":"0","path":"rust\/hgcli\/src\/main.rs","diffId":"8462","author":"kevincox","id":"54910","dateCreated":"1525593073","dateModified":"1525593073"},{"type":"accept","author":"kevincox","id":"54909","dateCreated":"1525593073","dateModified":"1525593073"},{"type":"update","diffId":"8462","author":"indygreg","id":"54834","dateCreated":"1525579874","dateModified":"1525579874"}],"dateCreated":"1525579874","dateModified":"1527793647","status":"Accepted"},{"id":"3446","callsign":"HG","title":"dispatch: move sys.exit() to caller (API)","author":"indygreg","summary":"Moving sys.exit() to the caller will make it easier to implement\nexit handling in Rust `hg` by eliminating the need to handle\nSystemExit in Rust.\n\n.. api::\n\n   dispatch.run() now returns an integer exit code instead of calling\n   sys.exit(). Callers of dispatch.run() must now exit.","testPlan":"","lineCount":"6","dependsOn":["3445"],"reviewers":["lothiraldan","durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"57755","dateCreated":"1527793642","dateModified":"1527793642"},{"type":"update","diffId":"8655","author":"indygreg","id":"56038","dateCreated":"1526100493","dateModified":"1526100493"},{"type":"accept","author":"lothiraldan","id":"54928","dateCreated":"1525677084","dateModified":"1525677084"},{"type":"update","diffId":"8461","author":"indygreg","id":"54824","dateCreated":"1525579873","dateModified":"1525579873"}],"dateCreated":"1525579873","dateModified":"1527793642","status":"Accepted"},{"id":"3663","callsign":"HG","title":"graft: add test showing --continue not preserving --edit passed earlier","author":"pulkit","summary":"Right now, if you run `hg graft --edit`, face some conflicts, resolve them and\nthen run `hg graft --continue`, you will see graft no longer pops up for the\neditor and you hate mercurial for not remembering arguments you passed earlier.\n\nUpcoming patch will fix this buggy behavior and this test will help us in\nrealizing those changes.","testPlan":"","lineCount":"41","dependsOn":[],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"8908","author":"pulkit","id":"57562","dateCreated":"1527287120","dateModified":"1527287120"}],"dateCreated":"1527287120","dateModified":"1527287308","status":"Needs Review"},{"id":"2875","callsign":"HG","title":"remotenames: introduce a config option to allow creation of remote bookmarks","author":"pulkit","summary":"This patch introduces another config option which if set to true, we can create\nnew bookmarks on the server using push command. An earlier patch added\ncapability to push to a certain bookmark on the server. If bookmark was not\nalready present on the server, we errored out. With\nremotenames.createremotebookmark config set, the push will create that bookmark\non the server.\n\nThis series serves as a great feature for people who use bookmarks as after this\nyou can push changesets to a bookmark which does not exists locally. In other\nwords, you can push your changesets to a bookmark without having it locally.\n\nThis series will help people in having a non-local bookmark workflow involving\nremotenames.\n\nThe functionality added in this patch is equivalent to --create flag to push\ncommand added by hgremotenames extension.","testPlan":"","lineCount":"166","dependsOn":[],"reviewers":["indygreg"],"ccs":["mercurial-devel"],"actions":[{"type":"accept","author":"indygreg","id":"52032","dateCreated":"1523468132","dateModified":"1523468132"},{"type":"update","diffId":"7288","author":"pulkit","id":"47564","dateCreated":"1522064875","dateModified":"1522064875"},{"type":"update","diffId":"7251","author":"pulkit","id":"47276","dateCreated":"1521704137","dateModified":"1521704137"},{"type":"update","diffId":"7105","author":"pulkit","id":"46425","dateCreated":"1521448583","dateModified":"1521448583"},{"type":"update","diffId":"7063","author":"pulkit","id":"46164","dateCreated":"1521121786","dateModified":"1521121786"}],"dateCreated":"1521121786","dateModified":"1526652340","status":"Accepted"},{"id":"2873","callsign":"HG","title":"remotenames: add functionality to override -B flag of push","author":"pulkit","summary":"This patch adds a new config option which can be used to override the `-B` flag\nof push command. If config is set to true, changesets will be pushed to that\nbookmark on the server. This is equivalent to `--to` flag of remotenames and\ninfinitepush.\n\nAfter this patch if config is set to true, the behavior will be:\n\n  * Normal push without -B flag: no behavior change\n  * specifying one bookmark using -B flag: changesets will be pushed to that\n        bookmark on the server, as explained above. Bookmark pushed will be on\n        the head most revision of the changesets pushed.\n  * specifying multiple bookmarks using -B flag: no behavior change, the same\n        behavior as it's today without this config option\n  * specifying multiple topologcial heads to push: no behvaior change\n\nThere needs some extra logic to update the remotenames state locally after the\npush since we know that the bookmark will now be at a known updated location.","testPlan":"","lineCount":"356","dependsOn":[],"reviewers":["indygreg"],"ccs":["smf","indygreg","mercurial-devel"],"actions":[{"type":"comment","comment":">>! In D2873#52100, @indygreg wrote:\n>>>! In D2873#52025, @smf wrote:\n>>>>! In D2873#52021, @indygreg wrote:\n>>> Looks good!\n>> \n>> I'm very heavily against this direction. Changing the behavior of push (even in this extension) is something I've always considered outside the scope of remotenames. Having another extension that changes push behavior (e.g. bookmark-push) is where I think this should go so that remotenames is just that: keeping track of remote names.\n> \n> I think there's room for this feature to live outside of remotenames. But currently I think it is the best place for it, since remotenames is the closest thing we have to... tracking remote names. We can always alias the old config option in the future if we move this functionality elsewhere.\n\nOne of the biggest regrets I have about the original remotenames, is the modifying of push (and some default) behavior. The `--to` is small enough (and I guess fine enough) for now but I absolutely and strongly believe that all the other behavior modifications should stay in an out-of-core repo for now. My goal is to split my remotenames repo to be based off of core's remotenames (basically only having the behavior changing patches there).","author":"smf","id":"52118","dateCreated":"1523482769","dateModified":"1523482769"},{"type":"comment","comment":">>! In D2873#52025, @smf wrote:\n>>>! In D2873#52021, @indygreg wrote:\n>> Looks good!\n> \n> I'm very heavily against this direction. Changing the behavior of push (even in this extension) is something I've always considered outside the scope of remotenames. Having another extension that changes push behavior (e.g. bookmark-push) is where I think this should go so that remotenames is just that: keeping track of remote names.\n\nI think there's room for this feature to live outside of remotenames. But currently I think it is the best place for it, since remotenames is the closest thing we have to... tracking remote names. We can always alias the old config option in the future if we move this functionality elsewhere.","author":"indygreg","id":"52100","dateCreated":"1523480112","dateModified":"1523480112"},{"type":"comment","comment":">>! In D2873#52021, @indygreg wrote:\n> Looks good!\n\nI'm very heavily against this direction. Changing the behavior of push (even in this extension) is something I've always considered outside the scope of remotenames. Having another extension that changes push behavior (e.g. bookmark-push) is where I think this should go so that remotenames is just that: keeping track of remote names.","author":"smf","id":"52026","dateCreated":"1523467971","dateModified":"1523467971"},{"type":"inline","comment":"Nit: \"exist\". This can be fixed in flight.","replyTo":null,"isNewFile":"1","line":"113","lineLength":"0","path":"hgext\/remotenames.py","diffId":"7286","author":"indygreg","id":"52023","dateCreated":"1523467769","dateModified":"1523467769"},{"type":"comment","comment":"Looks good!","author":"indygreg","id":"52022","dateCreated":"1523467769","dateModified":"1523467769"},{"type":"accept","author":"indygreg","id":"52021","dateCreated":"1523467769","dateModified":"1523467769"},{"type":"comment","comment":"A gentle reminder for reviewing this series. If this config option gets in, we can use it in infinitepush extension too where for now we have overridden `-B` flag in a hacky way.","author":"pulkit","id":"48737","dateCreated":"1522601487","dateModified":"1522601487"},{"type":"update","diffId":"7286","author":"pulkit","id":"47560","dateCreated":"1522064849","dateModified":"1522064849"},{"type":"inline","comment":"What happens if we're pushing multiple heads? I suspect this will choose a head\/revision arbitrarily - maybe depending on the `-r` arguments to `hg push`.\n\nI think we need to validate that the outgoing revs are in a single DAG head and we should then pick the rev that is the DAG head.\n\nPlease add test coverage for multiple `-r` arguments and `-r` arguments that resolve to multiple heads.","replyTo":null,"isNewFile":"1","line":"99","lineLength":"0","path":"hgext\/remotenames.py","diffId":"7249","author":"indygreg","id":"47382","dateCreated":"1521831360","dateModified":"1521831360"},{"type":"inline","comment":"Nit: what is `--create`?","replyTo":null,"isNewFile":"1","line":"101","lineLength":"0","path":"hgext\/remotenames.py","diffId":"7249","author":"indygreg","id":"47381","dateCreated":"1521831360","dateModified":"1521831360"},{"type":"inline","comment":"Strictly speaking, we should probably wrap `exchange._pushdiscoverybookmarks` so other extensions can get involved. But I think this is fine.","replyTo":null,"isNewFile":"1","line":"287","lineLength":"0","path":"hgext\/remotenames.py","diffId":"7249","author":"indygreg","id":"47380","dateCreated":"1521831360","dateModified":"1521831360"},{"type":"comment","comment":"I'm happy with this feature.\n\nBut the code needs work around multiple revisions\/heads before it can be queued.","author":"indygreg","id":"47379","dateCreated":"1521831360","dateModified":"1521831360"},{"type":"reject","author":"indygreg","id":"47378","dateCreated":"1521831360","dateModified":"1521831360"},{"type":"update","diffId":"7249","author":"pulkit","id":"47274","dateCreated":"1521704112","dateModified":"1521704112"},{"type":"update","diffId":"7103","author":"pulkit","id":"46423","dateCreated":"1521448561","dateModified":"1521448561"},{"type":"comment","comment":"I am not sure the functionality and config names should go to core or not, so I have introduced them in this extension. Also suggestions for better config names are welcome.","author":"pulkit","id":"46265","dateCreated":"1521212432","dateModified":"1521212432"},{"type":"update","diffId":"7061","author":"pulkit","id":"46148","dateCreated":"1521121763","dateModified":"1521121763"}],"dateCreated":"1521121763","dateModified":"1526652340","status":"Accepted"},{"id":"3565","callsign":"FBHGX","title":"sparse: delete now that it's in the core repo","author":"martinvonz","summary":"It was moved added to core repo in abd7dedbaa36 (sparse: vendor\nFacebook-developed extension, 2017-07-01).","testPlan":"","lineCount":"2706","dependsOn":[],"reviewers":["quark"],"ccs":["quark"],"actions":[{"type":"reject","author":"quark","id":"56456","dateCreated":"1526597213","dateModified":"1526597213"},{"type":"comment","comment":"I think there are still some differences. Namely the integration with `simplecache`, which is helpful for performance. Does the core version have something similar?","author":"quark","id":"56344","dateCreated":"1526429367","dateModified":"1526429592"},{"type":"update","diffId":"8703","author":"martinvonz","id":"56328","dateCreated":"1526428875","dateModified":"1526428875"}],"dateCreated":"1526428875","dateModified":"1526597213","status":"Needs Revision"},{"id":"2880","callsign":"HG","title":"bundle: add the possibility to bundle bookmarks (issue5792)","author":"lothiraldan","summary":"Also take the wlock when unbundling. It shouldn't have a big impact on\nperformance.","testPlan":"","lineCount":"143","dependsOn":[],"reviewers":["indygreg"],"ccs":["martinvonz","indygreg","pulkit","mercurial-devel"],"actions":[{"type":"inline","comment":"I tried adding such scenarios but failed to produce a divergence scenario, no idea why.\n\nI don't think I will have time to debug the issue before the freeze, so let's skip this changeset for 4.6","replyTo":"53793","isNewFile":"1","line":"37","lineLength":"26","path":"tests\/test-bundle-bookmarks.t","diffId":"7077","author":"lothiraldan","id":"54248","dateCreated":"1523950897","dateModified":"1523950897"},{"type":"inline","comment":"I think it should be the same thing as pulling from another repo. Try putting the same changes in another repo and pulling from it without giving the path a name (nothing in [paths] config). I haven't checked what suffix will be used on the divergent bookmark, but it probably makes sense to use the same here (perhaps it's just going to be \"D1@1\" as I guessed before).","replyTo":"53729","isNewFile":"1","line":"37","lineLength":"26","path":"tests\/test-bundle-bookmarks.t","diffId":"7077","author":"martinvonz","id":"53793","dateCreated":"1523724103","dateModified":"1523724103"},{"type":"inline","comment":"I have added such test and indeed this case is not handled right now. I'm not aware of how Mercurial handle this case when exchanging bookmarks, I will need to take a look at the code to find out.\n\nWhat is the correct behavior here? Change the existing bookmark or the bookmark from the bundle?","replyTo":"46275","isNewFile":"1","line":"37","lineLength":"26","path":"tests\/test-bundle-bookmarks.t","diffId":"7077","author":"lothiraldan","id":"53729","dateCreated":"1523698843","dateModified":"1523698843"},{"type":"update","diffId":"8263","author":"lothiraldan","id":"53727","dateCreated":"1523698706","dateModified":"1523698706"},{"type":"comment","comment":">>! In D2880#52559, @indygreg wrote:\n> What's the status of this patch? Is it still reviewable? If so, let's get a rebased version submitted, just in case things have changed.\n\nI need to update the patch according to @martinvoz comment.\n\nI was also waiting on your comment on https:\/\/phab.mercurial-scm.org\/D2880#46263, do you think we will keep the option once the bundlespec system is revamped? If yes, we can make the config option a non-experimental one. But if it will be dropped soon, I think keeping it experimental until it's dropped is preferable. What do you think?","author":"lothiraldan","id":"52583","dateCreated":"1523555228","dateModified":"1523555228"},{"type":"comment","comment":"What's the status of this patch? Is it still reviewable? If so, let's get a rebased version submitted, just in case things have changed.","author":"indygreg","id":"52560","dateCreated":"1523549760","dateModified":"1523549760"},{"type":"reject","author":"indygreg","id":"52559","dateCreated":"1523549760","dateModified":"1523549760"},{"type":"inline","comment":"Can we have a similar test case where we create divergence? Create a fork in the graph in the debugdrawdag call above. Let's say you have commit F that branches off of B, then do something like this:\n\n```\n$ hg bundle --all bundle\n$ hg strip --no-backup C\n$ hg bookmarks -f -r F D1\n$ hg unbundle -q bundle\n$ hg log -G -T '{desc} {bookmarks}\\n'\n  o  E\n  |\n  o  D D1 D2\n  |\n  o  C\n  |\n  |  o F D1@1\n  |\/\n  o  B\n  |\n  o  A A1\n```\n\nI don't know if \"@1\" is the appropriate divergence marker, but I can't think of a better one. I haven't even looked at your code to try to guess what it would be in practice (if it would work at all).","replyTo":null,"isNewFile":"1","line":"37","lineLength":"26","path":"tests\/test-bundle-bookmarks.t","diffId":"7077","author":"martinvonz","id":"46275","dateCreated":"1521218729","dateModified":"1521218729"},{"type":"inline","comment":"I was afraid we might have to delete this option once we have the discussion about bundle spec format with @indygreg. If that would be the case, I would feel more comfortable deleting an experimental option.\n\nIf we agreed on moving bundle-phases out of experimental, I guess we can do the same for bundle-bookmarks.\n\n@indygreg Do you think we would still have this option around once we have new-style bundle spec?","replyTo":"46262","isNewFile":"1","line":"422","lineLength":"0","path":"mercurial\/configitems.py","diffId":"7077","author":"lothiraldan","id":"46264","dateCreated":"1521208181","dateModified":"1521208181"},{"type":"inline","comment":"Any reason why we have all these config option as experimental? We talked at sprint about moving things out of experimental and we agreed on `experimental.bundle-phases`. This one sounds similar.","replyTo":null,"isNewFile":"1","line":"422","lineLength":"0","path":"mercurial\/configitems.py","diffId":"7077","author":"pulkit","id":"46262","dateCreated":"1521206865","dateModified":"1521206865"},{"type":"update","diffId":"7077","author":"lothiraldan","id":"46256","dateCreated":"1521206506","dateModified":"1521206506"}],"dateCreated":"1521206506","dateModified":"1523950897","status":"Needs Review"},{"id":"3304","callsign":"HG","title":"formatter: add support for -Tcbor","author":"indygreg","summary":"The implementation in the formatter is hopefully pretty\nstraightforward.\n\nUnlike other formats, we don't emit a list\/array or need to\nhandle \"joining\" elements between emitted items. That's because\nCBOR doesn't require any joining element: you just read the\nnext byte to see what type it is then decode it. You do that for\nas long as there are bytes to read.\n\nAs long as we aren't nesting formatters, the choice to omit\nan outer container for data seems acceptable. If nothing else,\nit will make it easier for decoders who don't support streaming\nto decode data without buffering the entire array first.\n\nI'm a bit unsure about add --cbor support to the f tool by importing\nMercurial modules. It does work in the test environment. But there's\nno demand module importer, etc. Unlike the existing JSON and pickle\nformatters, there isn't an implementation in the standard library\nfor us to use. So unless we want to forgo parsing the CBOR (this\ndoesn't seem great because the data isn't human readable), we'll\nneed to import at least the CBOR package from mercurial.*.\n\nThis commit also doesn't implement -Tcbor for changeset printing.\nI'll have to figure that out...\n\nI added tests for -Tcbor for most of the instances of -Tjson in\nthe test harness.\n\n.. feature::\n\n   `--template\/-T cbor` can now be specified to emit data as CBOR.\n\n   (This feature is experimental and the data format may change.)","testPlan":"","lineCount":"419","dependsOn":[],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"plan-changes","author":"indygreg","id":"53405","dateCreated":"1523664767","dateModified":"1523664767"},{"type":"update","diffId":"8105","author":"indygreg","id":"52788","dateCreated":"1523600792","dateModified":"1523600792"}],"dateCreated":"1523600792","dateModified":"1523664767","status":"Changes Planned"},{"id":"3209","callsign":"HG","title":"amend: exit 0 if there are no changes","author":"spectral","summary":"It's useful for writing some scripts or shell aliases if `hg amend` doesn't\nabort the series of commands if there's nothing to amend.  Users can then have\nsomething like \"alias amendevo='hg amend && hg evolve -a'\" to amend and evolve\nin one step.\n\nTo do this otherwise would require something like this:\n\n    alias amendevo='[[ -z \"$(hg status -mar)\" ]] && hg amend && hg evolve -a'\n\nWhile that's not too onerous, it's not immediately obvious to most that this is\nwhat would have to be done, so it's a small hurdle that users either have to\nspend time to figure out on their own or ask about.\n\nA similar change was made to `hg evolve` recently, for similar reasons.","testPlan":"","lineCount":"11","dependsOn":[],"reviewers":[],"ccs":["quark","mercurial-devel"],"actions":[{"type":"comment","comment":"I think it depends on what scripts want to do. I guess mpm's original point is, suppose you have a build script, or something that should do nothing if nothing changed, then `hg amend && build_script` just works as expected. If amend returns 0, then it'd be more complex to detect \"nothing changed\" case.","author":"quark","id":"51517","dateCreated":"1523325101","dateModified":"1523325101"},{"type":"comment","comment":">>! In D3209#51512, @quark wrote:\n> Returning 1 is actually more consistent with other core commands like pull, push, commit. See https:\/\/www.mercurial-scm.org\/pipermail\/mercurial-devel\/2012-January\/037711.html. Scripts should be updated to use `$?` explicitly. \n\nI'm not sure how scripts can be updated to use $? explicitly.  Mercurial exits 1 in many situations:\n- nothing found\n- hg verify, if there's any problems\n- unresolved conflicts when updating\/rebasing\/whatever\n- histedit 'edit' seems to dump you back into a terminal with exit status 1\n- MQ seems to use it a lot\n- I think a few others?\n\nYes, none of these apply to `hg amend`, so doing something like `hg amend; [[ $? -eq 1 ]] && hg evolve -a` is feasible, but again, feels really weird and non-discoverable. :\/\n\nI also see a few cases that return 2, mostly in MQ stuff (though there is one case in test-bookmarks-pushpull.t that's a little odd).  There's obviously the \"255\" return whenever there's a major problem.  \n\nI'm not too tied to the patch and can abandon it if people feel it's incorrect.  \n","author":"spectral","id":"51514","dateCreated":"1523319609","dateModified":"1523319609"},{"type":"comment","comment":"Returning 1 is actually more consistent with other core commands like pull, push, commit. See https:\/\/www.mercurial-scm.org\/pipermail\/mercurial-devel\/2012-January\/037711.html. Scripts should be updated to use `$?` explicitly. ","author":"quark","id":"51513","dateCreated":"1523317049","dateModified":"1523317049"},{"type":"update","diffId":"7916","author":"spectral","id":"51436","dateCreated":"1523311789","dateModified":"1523311789"}],"dateCreated":"1523311789","dateModified":"1523325101","status":"Needs Review"},{"id":"2403","callsign":"HG","title":"stack: remove destutil.stackbase","author":"lothiraldan","summary":"Use directly stack.getstack in show stack command. Remove destutil.stackbase.\nImport a TODO from the show stack in the code.","testPlan":"","lineCount":"16","dependsOn":["2402"],"reviewers":[],"ccs":["indygreg","mercurial-devel"],"actions":[{"type":"update","diffId":"7736","author":"lothiraldan","id":"50324","dateCreated":"1522945351","dateModified":"1522945351"},{"type":"comment","comment":">>! In D2403#39460, @indygreg wrote:\n> Oh, I guess the series never did formerly define a revset for stack :\/\n> \n> I think there is room for one. We have the ability to mark revsets as experimental or internal, right? Could\/should we do that so we don't need `hg debugstack`?\n> \n> Anyway, I could probably queue this with some minor rework. But I'm going to hold off a bit and see if others have opinions.\n\nI have updated the series without using debugstack, thank you for the suggestion. I thought revset would were much harder to implement but turn out not so much harder.","author":"lothiraldan","id":"40956","dateCreated":"1519934450","dateModified":"1519934450"},{"type":"update","diffId":"6219","author":"lothiraldan","id":"40701","dateCreated":"1519862380","dateModified":"1519862380"},{"type":"comment","comment":"Oh, I guess the series never did formerly define a revset for stack :\/\n\nI think there is room for one. We have the ability to mark revsets as experimental or internal, right? Could\/should we do that so we don't need `hg debugstack`?\n\nAnyway, I could probably queue this with some minor rework. But I'm going to hold off a bit and see if others have opinions.","author":"indygreg","id":"39461","dateCreated":"1519440298","dateModified":"1519440298"},{"type":"update","diffId":"6017","author":"lothiraldan","id":"39366","dateCreated":"1519382238","dateModified":"1519382238"}],"dateCreated":"1519382238","dateModified":"1522945387","status":"Needs Review"},{"id":"2402","callsign":"HG","title":"stack: introduce an option to limit the selection on the current branch","author":"lothiraldan","summary":" ","testPlan":"","lineCount":"37","dependsOn":["2401"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"7735","author":"lothiraldan","id":"50323","dateCreated":"1522945345","dateModified":"1522945345"},{"type":"update","diffId":"6218","author":"lothiraldan","id":"40700","dateCreated":"1519862377","dateModified":"1519862377"},{"type":"update","diffId":"6016","author":"lothiraldan","id":"39357","dateCreated":"1519382234","dateModified":"1519382234"}],"dateCreated":"1519382234","dateModified":"1522945345","status":"Needs Review"},{"id":"2401","callsign":"HG","title":"stack: introduce an option to disable the restriction on ancestor","author":"lothiraldan","summary":"The default stack revset restrict on only the ancestors of a revision. Some\ntools might want to display the current stack even if the working directory\nparent is in the middle of the stack.\n\nAdd the `stack.restrict-ancestors` option to disable that.","testPlan":"","lineCount":"17","dependsOn":["2400"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"7734","author":"lothiraldan","id":"50322","dateCreated":"1522945341","dateModified":"1522945341"},{"type":"update","diffId":"6217","author":"lothiraldan","id":"40699","dateCreated":"1519862374","dateModified":"1519862374"},{"type":"update","diffId":"6015","author":"lothiraldan","id":"39347","dateCreated":"1519382229","dateModified":"1519382229"}],"dateCreated":"1519382229","dateModified":"1522945341","status":"Needs Review"},{"id":"2400","callsign":"HG","title":"stack: begin to make the stack revset configurable","author":"lothiraldan","summary":"First extrat the not-merge part in its own configuration knob `stack.not-\nmerge`.","testPlan":"","lineCount":"25","dependsOn":[],"reviewers":["indygreg"],"ccs":["indygreg","mercurial-devel"],"actions":[{"type":"update","diffId":"7733","author":"lothiraldan","id":"50320","dateCreated":"1522945336","dateModified":"1522945336"},{"type":"comment","comment":"The goal of this series is to define a clean API so that commands and tools can be built in a stack-definition agnostic way. It's not to define a one-size-fits-all stack definition.\n\nPeople and companies have different workflows and needs. Their definition of a stack will likely to be different to better match their workflow. I tried finding a way to configure stack as easily as possible. I have the impression that using configuration options is slightly easier than defining a custom extension. I also avoided using directly a revset from the configuration file as I'm not sure it would be the best way to express all situations.\n\nI don't know if the config options best place are in the `stack` section though.","author":"lothiraldan","id":"50310","dateCreated":"1522943035","dateModified":"1522943035"},{"type":"comment","comment":"Reading through the remainder of the series, I like the customization of stack evaluation. But the use of config options seems weird to me. It isn't clear why we need config options. It feels like these options should be passed in as function arguments. Now, it will be useful to test these variations. Maybe that does require config options. But until there is a user-facing need for these config options, It feels better to put them in `[devel]` or something like that.\n\nIs there a future consumer of these config options that justifies them being normal config options?","author":"indygreg","id":"46534","dateCreated":"1521490461","dateModified":"1521490461"},{"type":"reject","author":"indygreg","id":"46533","dateCreated":"1521490461","dateModified":"1521490461"},{"type":"inline","comment":"As I review just this patch, I question if we shouldn't just concatenate `and not ::merge()` to the revset string. But I know other changes are coming. I assume the final state justifies the extra complexity.","replyTo":null,"isNewFile":"1","line":"41","lineLength":"1","path":"mercurial\/stack.py","diffId":"6216","author":"indygreg","id":"46531","dateCreated":"1521490117","dateModified":"1521490117"},{"type":"comment","comment":"I'm not super keen on introducing a new `[stack]` config section. But I don't have any better ideas for where this should go. `[ui]` is my best idea, but I thought we decided we wanted to stop stuffing random stuff into `[ui]`. So I dunno.","author":"indygreg","id":"46530","dateCreated":"1521490117","dateModified":"1521490117"},{"type":"accept","author":"indygreg","id":"46529","dateCreated":"1521490117","dateModified":"1521490117"},{"type":"update","diffId":"6216","author":"lothiraldan","id":"40698","dateCreated":"1519862372","dateModified":"1519862372"},{"type":"update","diffId":"6014","author":"lothiraldan","id":"39337","dateCreated":"1519382224","dateModified":"1519382224"}],"dateCreated":"1519382224","dateModified":"1522945336","status":"Needs Review"},{"id":"2999","callsign":"HG","title":"infinitepush: use zope.interface to define indexapi interface","author":"pulkit","summary":"This patch makes indexapi use zope.interface which is now vendored with\nmercurial.","testPlan":"","lineCount":"20","dependsOn":[],"reviewers":["indygreg"],"ccs":["indygreg","mercurial-devel"],"actions":[{"type":"comment","comment":"Oh, when you add test coverage, watch out for `mysql.connector` being unavailable. You may want to register a fake module in `sys.modules['mysql.connector']` to ensure it always is available.","author":"indygreg","id":"48566","dateCreated":"1522517045","dateModified":"1522517045"},{"type":"inline","comment":"`Interface` classes don't define `self` on methods. So please drop `self` from all these methods.","replyTo":null,"isNewFile":"1","line":"25","lineLength":"0","path":"hgext\/infinitepush\/indexapi.py","diffId":"7466","author":"indygreg","id":"48564","dateCreated":"1522516965","dateModified":"1522516965"},{"type":"comment","comment":"Please also add test coverage to `test-check-interfaces.py`, otherwise these interfaces mean practically nothing. `zope.interface` doesn't perform run-time interface conformance tests unless you explicitly instruct it to. For now, we're going to verify interfaces in `test-check-interfaces.py`.","author":"indygreg","id":"48563","dateCreated":"1522516965","dateModified":"1522516965"},{"type":"reject","author":"indygreg","id":"48562","dateCreated":"1522516965","dateModified":"1522516965"},{"type":"update","diffId":"7466","author":"pulkit","id":"48513","dateCreated":"1522497241","dateModified":"1522497241"}],"dateCreated":"1522497241","dateModified":"1522517045","status":"Needs Revision"},{"id":"2623","callsign":"HG","title":"dispatch: adding config items for overriding flag defaults","author":"rdamazio","summary":"This introduces the new defaults format \"commandname.default.optionname\" which\ndirectly overrides the default of the option, instead of prepending the\ncommand-line option. This is meant to replace the [defaults] section which is\nalready deprecated in a manner that's easier and safer to use than creating\naliases.","testPlan":"","lineCount":"96","dependsOn":["2622"],"reviewers":["yuja"],"ccs":["dploch","mercurial-devel"],"actions":[{"type":"inline","comment":"I thought callables were meant to be used to generate the default default, not with overridden values?","replyTo":"43409","isNewFile":"1","line":"625","lineLength":"1","path":"mercurial\/dispatch.py","diffId":"6609","author":"rdamazio","id":"47743","dateCreated":"1522132208","dateModified":"1522132208"},{"type":"inline","comment":"Nobody should use the same *instance* on multiple flags. Even with the current flags, if you use e.g. the same list on many, that'll cause problems with listopt.\n","replyTo":"43410","isNewFile":"1","line":"639","lineLength":"1","path":"mercurial\/dispatch.py","diffId":"6609","author":"rdamazio","id":"47742","dateCreated":"1522132208","dateModified":"1522132208"},{"type":"inline","comment":"This makes me nervous.  What if someone re-uses a customopt instance in multiple commands?  i.e.:\n\nDATE_FLAG = mypkg.dateopt()\n...\n('b', 'before', DATE_FLAG, '')\n('a', 'after', 'DATE_FLAG', '')\n\nNow, setting commands.defaults.before=2018-03-05 also silently changes the default for 'after'.  I suspect we need to introduce a wrapper class like what I suggest on lines 625-625, that delegates and leaves the original default unchanged.  And either way, we should probably clarify in the docs on customopts what expected use of the class is (i.e., should we just forbid reuse, is 'oldstate' safe to mutate, etc.)","replyTo":null,"isNewFile":"1","line":"639","lineLength":"1","path":"mercurial\/dispatch.py","diffId":"6609","author":"dploch","id":"43410","dateCreated":"1520292544","dateModified":"1520292544"},{"type":"inline","comment":"This doesn't handle callables properly.  I wonder if the something like the following would work instead:\n\noldopt = fancyopts._defaultopt(olddefault)\nnewdefault = old.opt.newstate(olddefault, ui.config(\"commands\", cfgitem)\nc[idx] = (opt[0], opt[1], fancyopts._withnewdefault(oldopt, newdefault), opt[3])\n\nWhere '_withnewdefault' is a wrapper customopt that just changes the default.","replyTo":null,"isNewFile":"1","line":"625","lineLength":"1","path":"mercurial\/dispatch.py","diffId":"6609","author":"dploch","id":"43409","dateCreated":"1520292544","dateModified":"1520292544"},{"type":"inline","comment":"I think rdamazio's response is correct.  The type information for an opt exists only in the 'defaultvalue' slot in the tuple, so it needs to be extracted from there before reading the config.  I don't have a strong opinion as to where this code should go - 'ui' does seem marginally more appropriate than 'fancyopts', since it keeps 'fancyopts' low-level and not depending on 'ui'.","replyTo":"42582","isNewFile":"1","line":"624","lineLength":"0","path":"mercurial\/dispatch.py","diffId":"6523","author":"dploch","id":"43408","dateCreated":"1520292544","dateModified":"1520292544"},{"type":"inline","comment":"Ah, makes sense. See if this addresses that case satisfactorily - it still has the caveat of not being able to \"reset\" container types, but that's true of the command line as well (if you have a list flag with a non-empty default, there's no way to remove that default item).","replyTo":"42770","isNewFile":"1","line":"624","lineLength":"0","path":"mercurial\/dispatch.py","diffId":"6523","author":"rdamazio","id":"43071","dateCreated":"1520195950","dateModified":"1520195950"},{"type":"update","diffId":"6609","author":"rdamazio","id":"43064","dateCreated":"1520195818","dateModified":"1520195818"},{"type":"inline","comment":"IIUC, an extension author may implement its own customopt subclasses, and\nput them into the command table, so we can't make ui.configtyped to\nsupport all of them.","replyTo":"42686","isNewFile":"1","line":"624","lineLength":"0","path":"mercurial\/dispatch.py","diffId":"6523","author":"yuja","id":"42770","dateCreated":"1520183736","dateModified":"1520183736"},{"type":"inline","comment":"You're right, I had the plain logic inverted in my head. Removed.","replyTo":"42581","isNewFile":"1","line":"390","lineLength":"0","path":"mercurial\/ui.py","diffId":"6523","author":"rdamazio","id":"42687","dateCreated":"1520181261","dateModified":"1520181261"},{"type":"inline","comment":"The issue is that customopt (and all its children) assume the value type is already the correct one, and thus do not perform any conversion. Since we're parsing values from the config file, the conversion is desired to ensure they don't all end up as text - the config{bool,int,etc} methods called by configtyped perform the proper conversion. In most cases (all commands that declare default values) no conversio is needed since those already have the correct type.\n","replyTo":"42582","isNewFile":"1","line":"624","lineLength":"0","path":"mercurial\/dispatch.py","diffId":"6523","author":"rdamazio","id":"42686","dateCreated":"1520181261","dateModified":"1520181261"},{"type":"update","diffId":"6548","author":"rdamazio","id":"42684","dateCreated":"1520181250","dateModified":"1520181250"},{"type":"inline","comment":"Maybe this type conversion can be a `fancyopt.customopt` method since we've\nrefactored the default handling by D2090?\n\n```\n# no idea if _defaultopt() should be made public or the whole commands.default handling\n# should be moved to fancyopts\nx = fancyopts._defaultopt(olddefault)\nnewdefault = x.configdefault(ui, cmd, optname, ...)\n```\n\n@dploch, any suggestions?","replyTo":null,"isNewFile":"1","line":"624","lineLength":"0","path":"mercurial\/dispatch.py","diffId":"6523","author":"yuja","id":"42582","dateCreated":"1520173297","dateModified":"1520173297"},{"type":"inline","comment":"Perhaps this is noop since `[commands]` is removed at all if `ui.plain()` returns True.","replyTo":null,"isNewFile":"1","line":"390","lineLength":"0","path":"mercurial\/ui.py","diffId":"6523","author":"yuja","id":"42581","dateCreated":"1520173297","dateModified":"1520173297"},{"type":"reject","author":"yuja","id":"42578","dateCreated":"1520173297","dateModified":"1520173297"},{"type":"comment","comment":"FYI this is a change I had previously sent to the list as 60b3222e01f96f91ece7eda9681a89bf3bb930a6, and Yuya reviewed . I just had never followed up on it.","author":"rdamazio","id":"42436","dateCreated":"1520122245","dateModified":"1520122245"},{"type":"update","diffId":"6523","author":"rdamazio","id":"42435","dateCreated":"1520121514","dateModified":"1520121514"},{"type":"update","diffId":"6521","author":"rdamazio","id":"42421","dateCreated":"1520121282","dateModified":"1520121282"}],"dateCreated":"1520121282","dateModified":"1522132208","status":"Needs Review"},{"id":"2057","callsign":"HG","title":"rust implementation of hg status","author":"Ivzhh","summary":"* implementation of revlog v1\n* parsing changelog, manifest, dirstate\n* use .hgignore in repo root\n* comparable performance with current hg status (Linux & Mac: slightly faster, Windows: slightly slower)\n* use hg r-status as subcommand, in this case, bypass python engine","testPlan":"","lineCount":"3066","dependsOn":[],"reviewers":["kevincox"],"ccs":["quark","yuja","glandium","krbullock","indygreg","durin42","kevincox","mercurial-devel"],"actions":[{"type":"inline","comment":"I meant safe not as it it didn't need the unsafe keyword, but in that the use of the `unsafe` block is safe.\n\nIt should really be called the `trust_me,_I_know_this_is_safe` block. But since you are not getting the compiler checking it is often useful to add a comment why the action you are performing is correct. In this case it is correct because the caller initializes this variable before the function is called.","replyTo":"46950","isNewFile":"1","line":"91","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"47428","dateCreated":"1521888284","dateModified":"1521888284"},{"type":"inline","comment":"Actually I didn't realize that RwLock doesn't get a regular `get()` since it is doing a compile time borrow check. https:\/\/doc.rust-lang.org\/std\/sync\/struct.RwLock.html#method.get_mut. My mistake, the code is fine.","replyTo":"46960","isNewFile":"1","line":"136","lineLength":"0","path":"rust\/hgstorage\/src\/local_repo.rs","diffId":"6724","author":"kevincox","id":"47427","dateCreated":"1521888284","dateModified":"1521888284"},{"type":"inline","comment":"Does it make sense to make `DirStateEntry.mtime` be a `std::time::SystemTime` and convert upon reading the structure in?\n\nIf not I would prefer doing the conversion here:\n\n```\nelse if mtd.modified().unwrap() == UNIX_EPOCH + Duration::from_secs(dir_entry.mtime as u64) {\n```\n\n(Maybe extract the system time to higher up, or even a helper function on dir_entry)","replyTo":null,"isNewFile":"1","line":"263","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"7188","author":"kevincox","id":"47426","dateCreated":"1521888284","dateModified":"1521888284"},{"type":"inline","comment":"Sorry, I misunderstood the logic. You can do this:\n\n```\ndiff -r ccc683587fdb rust\/hgstorage\/src\/dirstate.rs\n--- a\/rust\/hgstorage\/src\/dirstate.rs\tSat Mar 24 10:05:53 2018 +0000\n+++ b\/rust\/hgstorage\/src\/dirstate.rs\tSat Mar 24 10:14:58 2018 +0000\n@@ -184,8 +184,7 @@\n                         continue;\n                     }\n \n-                    if self.dir_state_map.contains_key(rel_path) {\n-                        let dir_entry = &self.dir_state_map[rel_path];\n+                    if let Some(dir_entry) = self.dir_state_map.get(rel_path) {\n                         files_not_in_walkdir.remove(rel_path);\n                         DirState::check_status(&mut res, abs_path, rel_path, dir_entry);\n                     } else if !matcher.check_path_ignored(rel_path.to_str().unwrap()) {\n```","replyTo":"46959","isNewFile":"1","line":"170","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"47425","dateCreated":"1521888284","dateModified":"1521888284"},{"type":"inline","comment":"Switch the return type to `std::io::Result` and then you can have\n\n```\nlet metadata = p.metadata()?;\nlet mtime = metadata.modified()?;\n\/\/ ...\n```","replyTo":null,"isNewFile":"1","line":"103","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"7188","author":"kevincox","id":"47424","dateCreated":"1521888284","dateModified":"1521888284"},{"type":"inline","comment":"A link to the mentioned wiki page would be very helpful to readers.","replyTo":null,"isNewFile":"1","line":"50","lineLength":"0","path":"rust\/hgstorage\/src\/config.rs","diffId":"7188","author":"kevincox","id":"47423","dateCreated":"1521888284","dateModified":"1521888284"},{"type":"inline","comment":"These are `HashSet`'s which don't have a defined iterator order. IIRC the python implementation sorts the results which is probably desirable.","replyTo":null,"isNewFile":"1","line":"260","lineLength":"0","path":"rust\/hgcli\/src\/main.rs","diffId":"7188","author":"kevincox","id":"47422","dateCreated":"1521888284","dateModified":"1521888284"},{"type":"inline","comment":"If this computation only depends on `len` it would be nice to put it in a helper function.","replyTo":null,"isNewFile":"1","line":"105","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"7188","author":"kevincox","id":"47421","dateCreated":"1521888284","dateModified":"1521888284"},{"type":"inline","comment":"I get that, but I still think it makes the code easier to read when the python-interop and the logic as separated where it is easy to do so.","replyTo":"46947","isNewFile":"1","line":"23","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"47420","dateCreated":"1521888284","dateModified":"1521888284"},{"type":"comment","comment":"The latest changes are looking really good. I have a couple more comments but I didn't have time for a full review. I'll try to get more reviewed tomorrow. It seems that you still have a lot of stuff still in-flight so I'll try to slowly review the changes as I have time. If you want input\/feedback on any particular part just ask and I will prioritize it.\n\nThis change is very large so it might be worth splitting off a smaller component and getting that submitted first. However I do realize that for starting out it is often helpful to get some actual use cases implemented before committing the base structures.","author":"kevincox","id":"47419","dateCreated":"1521888284","dateModified":"1521888284"},{"type":"comment","comment":">>! In D2057#46980, @Ivzhh wrote:\n>>>! In D2057#46726, @yuja wrote:\n>>>> I think the only place where you would need to do os-specific code is when\n>>>> doing serialization and serialization\n>>> \n>>> Yes, that will be feasible in strictly typed language like Rust.\n>> \n>> To be clear, I meant serialization\/deserialization between filesystem path and\n>> internal dirstate\/manifest path, not between dirstate storage and in-memory\n>> dirstate object.\n> \n> I guess your suggestion is like this: @yuja\n> \n> 1. if it is windows and the code page is MBCS, try to decode the paths read from manifest and dirstate into unicode equivalent\n> 2. use utf internally and with rust IO api\n> 3. when writing back to dirstate and manifest, encode utf to MBCS\n\nNo. My suggestion is:\n\n1. keep manifest\/dirstate paths as bytes (which are probably wrapped by some type, say HgPath)\n2. but we want to use Rust's standard library for I\/O\n3. so, add utility function\/trait to convert HgPath to Path\/PathBuf, where MBCS-Wide conversion\n   will occur.\n\nI think raw byte paths will be needed to build store paths (e.g. `.hg\/store\/data\/~2eclang-format.i`).\n\nhttps:\/\/www.mercurial-scm.org\/repo\/hg\/file\/4.5.2\/mercurial\/store.py","author":"yuja","id":"47287","dateCreated":"1521730814","dateModified":"1521730814"},{"type":"comment","comment":">>! In D2057#46726, @yuja wrote:\n>>> I think the only place where you would need to do os-specific code is when\n>>> doing serialization and serialization\n>> \n>> Yes, that will be feasible in strictly typed language like Rust.\n> \n> To be clear, I meant serialization\/deserialization between filesystem path and\n> internal dirstate\/manifest path, not between dirstate storage and in-memory\n> dirstate object.\n\nI guess your suggestion is like this: @yuja\n\n1. if it is windows and the code page is MBCS, try to decode the paths read from manifest and dirstate into unicode equivalent\n2. use utf internally and with rust IO api\n3. when writing back to dirstate and manifest, encode utf to MBCS\n\nPlease let me know if I have misunderstanding. Thank you!","author":"Ivzhh","id":"46981","dateCreated":"1521662441","dateModified":"1521662461"},{"type":"inline","comment":"I guess I did this because I met some empty change delta in the beginning. I think I won't try to parallelize unzip for now.","replyTo":"43866","isNewFile":"1","line":"290","lineLength":"3","path":"rust\/hgstorage\/src\/revlog_v1.rs","diffId":"6724","author":"Ivzhh","id":"46964","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"I think it explains why in mercurial repo, rust version is significantly faster. I am working on cpu future, but I did not finalize design style yet. I will keep working on that.","replyTo":"43867","isNewFile":"1","line":"279","lineLength":"0","path":"rust\/hgstorage\/src\/revlog_v1.rs","diffId":"6724","author":"Ivzhh","id":"46963","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"This part, including the stream-style, is from python part. I will update later with xi-rope.","replyTo":"43969","isNewFile":"1","line":"51","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"Ivzhh","id":"46962","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"I borrow this logic as whole from python code. It will need sometime to re-translate to non-pointer-arithmetic way.","replyTo":"43949","isNewFile":"1","line":"14","lineLength":"0","path":"rust\/hgstorage\/src\/matcher.rs","diffId":"6724","author":"Ivzhh","id":"46961","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"I think LRU will update reference count (or timestamp) when the data is accessed.","replyTo":"43942","isNewFile":"1","line":"136","lineLength":"0","path":"rust\/hgstorage\/src\/local_repo.rs","diffId":"6724","author":"Ivzhh","id":"46960","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"I kind of get borrow check compile error here. Later I use Occupied() when possible.","replyTo":"43930","isNewFile":"1","line":"170","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"Ivzhh","id":"46959","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"I add the error handling back","replyTo":"43923","isNewFile":"1","line":"161","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"Ivzhh","id":"46958","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"For the filter, I follow the example in the walkdir doc. I guess what I want is to skip the dir for later recursive visiting. ","replyTo":"43921","isNewFile":"1","line":"152","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"Ivzhh","id":"46957","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"I think dir state needs to 1. read existing one; 2. create one if not exits; maybe private for now.","replyTo":"43916","isNewFile":"1","line":"108","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"Ivzhh","id":"46956","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"I remember the python hg uses the name, in the beginning, I tried to replicate py-hg's behaviour. But I think it needs to be renamed. I agree with you.","replyTo":"43928","isNewFile":"1","line":"48","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"Ivzhh","id":"46955","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"I like to_owned(), I will them in later occasions. Thank you!","replyTo":"43913","isNewFile":"1","line":"78","lineLength":"0","path":"rust\/hgstorage\/src\/config.rs","diffId":"6724","author":"Ivzhh","id":"46954","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"Sure, I will use v1 only for now. In the beginning I kinda over designed this part.","replyTo":"43865","isNewFile":"1","line":"95","lineLength":"0","path":"rust\/hgstorage\/src\/config.rs","diffId":"6724","author":"Ivzhh","id":"46953","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"I guess I will use the rest info later. hg seems put some meta data in the commit comments. I will keep it for now. Thank you!","replyTo":"43911","isNewFile":"1","line":"31","lineLength":"0","path":"rust\/hgstorage\/src\/changelog.rs","diffId":"6724","author":"Ivzhh","id":"46952","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"Thank you for the suggestion! I guess I need to extend clap later to support hg style command line. Right now whenever clap cannot handle the argument parsing, I will redirect the arguments to hg directly.","replyTo":"43864","isNewFile":"1","line":"233","lineLength":"28","path":"rust\/hgcli\/src\/main.rs","diffId":"6724","author":"Ivzhh","id":"46951","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"when I removed the unsafe, I got error: error[E0133]: use of mutable static requires unsafe function or block","replyTo":"43908","isNewFile":"1","line":"91","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"Ivzhh","id":"46950","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"I guess it is because NLL. When I started the work, rust compiler reported borrow check error on this part. I later read an article talking about NLL update in rust. But before that, I use the braces to avoid the error.","replyTo":"43901","isNewFile":"1","line":"46","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"Ivzhh","id":"46949","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"pad is a bool, however when I checked it in hg-python, int are passed to the function. I guess I need to update cpython wrapper for this, a more broad logic conversion.","replyTo":"43900","isNewFile":"1","line":"23","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"Ivzhh","id":"46948","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"This crate is my previous try to integrate rust into hg. Right now I guess mine main pursue is to add hg r-* commands for rust. I will follow your suggestion when I am implementing the wire protocol and reuse the code for pure rust crate.","replyTo":"43897","isNewFile":"1","line":"23","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"Ivzhh","id":"46947","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"inline","comment":"It should be any &[u8], but the current cpython crate doesn't wrap for &[u8]. I think I need to fork and add that part. I put it in my checklist now.","replyTo":"43898","isNewFile":"1","line":"23","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"Ivzhh","id":"46946","dateCreated":"1521658541","dateModified":"1521658541"},{"type":"comment","comment":"  - add revlog and mpatch facilities\n  - add changelog parsing\n  - add manifest parsing\n  - path encoding for data store\n  - add dirstate and matcher facilities\n  - add local repository and the supporting modules\n  - use cargo fmt to format code\n  - add hg r-status command\n  - bincode 1.0.0 is a bit slow in my test\n  - delay pattern matching during dir walk\n  - optimize out trie and enable CoreXL profiling\n  - use hashmap\n  - remove thread pool\n  - rust default read is not buf-ed, this is the key of slowness\n  - change to globset\n  - convert glob to regex\n  - hg ignore patterns are all converted to regex (as hg does), and now it is faster\n  - filter dir early to prevent walking\n  - Update matcher mod after testing Mozilla unified repo\n  - bug fix: use byte literals instead of numbers\n  - hg store path encoding is per byte style, update code according to Kevin Cox's comments\n  - update matcher testing according to Match interface change\n  - If clap fails to recognize r-* subcommands, then run python-version hg\n  - changelog coding style revised\n  - remove legacy revlog v0 and unfinished v2.\n  - partially revise the dirstate reviews\n  - remove duplicated build.rs, let the executable module guarantee the python\n  - use cursor in base85 encoding, reducing raw index-math\n  - use cursor in base85 decoding, reducing raw index-math\n  - dirstate update according to review comments\n  - config update according to review comments\n  - mpatch rename to more meaningful names\n  - simplify matcher as when there is no syntax named in the beginning, use regexp\n  - local repo coding style update\n  - dirstate coding style update\n  - manifest coding style update\n","author":"Ivzhh","id":"46944","dateCreated":"1521656608","dateModified":"1521656608"},{"type":"update","diffId":"7188","author":"Ivzhh","id":"46942","dateCreated":"1521656608","dateModified":"1521656608"},{"type":"comment","comment":"Ah, I forgot to consider the python interop. Now the need for that crate makes sense. Thanks for explaining.","author":"kevincox","id":"46733","dateCreated":"1521627628","dateModified":"1521627628"},{"type":"comment","comment":">> I think the only place where you would need to do os-specific code is when\n>> doing serialization and serialization\n> \n> Yes, that will be feasible in strictly typed language like Rust.\n\nTo be clear, I meant serialization\/deserialization between filesystem path and\ninternal dirstate\/manifest path, not between dirstate storage and in-memory\ndirstate object.","author":"yuja","id":"46726","dateCreated":"1521613817","dateModified":"1521613817"},{"type":"comment","comment":"> I think the only place where you would need to do os-specific code is when\n> doing serialization and serialization\n\nYes, that will be feasible in strictly typed language like Rust.\n\n> which I think should be handled by https:\/\/doc.rust-lang.org\/std\/os\/unix\/ffi\/trait.OsStringExt.html\n> and https:\/\/doc.rust-lang.org\/std\/os\/windows\/ffi\/trait.OsStringExt.html.\n\nNot true for Windows because Rust uses Unicode (UTF-16-ish) API, whereas\nPython 2 does ANSI. We need to convert a \"wide\" string to a locale-dependent string.\n\nMaybe the local-encoding crate will do that for us?\n\n","author":"yuja","id":"46723","dateCreated":"1521607475","dateModified":"1521607475"},{"type":"comment","comment":"I'm not a windows expert but it seems like the rust OsStr, Path and filesystem APIs should handle these conversions for you. I think the only place where you would need to do os-specific code is when doing serialization and serialization which I think should be handled by https:\/\/doc.rust-lang.org\/std\/os\/unix\/ffi\/trait.OsStringExt.html and https:\/\/doc.rust-lang.org\/std\/os\/windows\/ffi\/trait.OsStringExt.html.","author":"kevincox","id":"46685","dateCreated":"1521577349","dateModified":"1521577349"},{"type":"comment","comment":"https:\/\/crates.io\/crates\/local-encoding seems to be the right choice.","author":"quark","id":"46681","dateCreated":"1521573831","dateModified":"1521573831"},{"type":"comment","comment":"> I am looking at Mozilla's rust winapi bindings, let me see if I can directly wrap around [winapi::um::fileapi::FindFirstFileA](https:\/\/docs.rs\/winapi\/*\/x86_64-pc-windows-msvc\/winapi\/um\/fileapi\/fn.FindFirstFileA.html)\n\nThat's probably a hard way. I was thinking of something converting\nbetween OsStr (i.e. Path) and MBCS bytes by using Win32 API, instead\nof calling out the \"A\" API.\n\nhttps:\/\/msdn.microsoft.com\/en-us\/library\/windows\/desktop\/dd319072(v=vs.85).aspx\n\nWe don't do that in Python, but Rust's type system will help making it right.","author":"yuja","id":"44688","dateCreated":"1520680806","dateModified":"1520680806"},{"type":"comment","comment":">>! In D2057#44269, @yuja wrote:\n>>> Reading that page it seems to claim that filenames should be utf8, not bytes. If utf8, this is what the code does, but if it is bytes that definitely won't work.\n>> \n>> IIRC it's bytes everyplace except Windows, where we pretend utf8 is real?\n> \n> It's MBCS (i.e. ANSI multi-byte characters) on Windows. The plain was to support\n> both MBCS and UTF-8-variant on Windows, but that isn't a thing yet.\n> \n> Perhaps we'll have to write a platform compatibility layer (or serialization\/deserialization\n> layer) on top of the Rust's file API, something like vfs.py we have in Python code.\n\nThank you for confirming that, I am a bit confusing when I read Encoding Plan wiki page. I am looking at Mozilla's rust winapi bindings, let me see if I can directly wrap around [winapi::um::fileapi::FindFirstFileA](https:\/\/docs.rs\/winapi\/*\/x86_64-pc-windows-msvc\/winapi\/um\/fileapi\/fn.FindFirstFileA.html)","author":"Ivzhh","id":"44295","dateCreated":"1520618349","dateModified":"1520618349"},{"type":"comment","comment":">> Reading that page it seems to claim that filenames should be utf8, not bytes. If utf8, this is what the code does, but if it is bytes that definitely won't work.\n> \n> IIRC it's bytes everyplace except Windows, where we pretend utf8 is real?\n\nIt's MBCS (i.e. ANSI multi-byte characters) on Windows. The plain was to support\nboth MBCS and UTF-8-variant on Windows, but that isn't a thing yet.\n\nPerhaps we'll have to write a platform compatibility layer (or serialization\/deserialization\nlayer) on top of the Rust's file API, something like vfs.py we have in Python code.","author":"yuja","id":"44270","dateCreated":"1520598781","dateModified":"1520598781"},{"type":"comment","comment":"Hi everyone,\n\nThank you for your encouragements and comments! I will follow up with all comments and update the code soon.\n\n@indygreg It is a great idea to test on Mozilla repo, actually I found several things interesting:\n\n1. I found a bug in my code (shame on me): because I did not use byte literal, and I made a typo. This triggers problem in Mozilla unified repo\n2. A regexp pattern in hgignore in Mozilla unified repo is not supported by rust's regex crate, a.k.a. \"(?!)\". I choose to ignore these unsupported patterns.\n3. My version is slower in this repo: 70s (hg) and 90s (mine). CodeXL reveals that the mpatch::collect() function uses 63% of the running time. I think I need to optimize it somehow.\n\nI totally agree with @kevincox that I did not sort well on char\/u8\/str\/String\/Path\/PathBuf. The first bug is caused by this. I need to improve them.\n\nThank you everyone!","author":"Ivzhh","id":"44180","dateCreated":"1520558265","dateModified":"1520558265"},{"type":"comment","comment":"Rust has platform independent types [[ https:\/\/doc.rust-lang.org\/stable\/std\/path\/struct.PathBuf.html | `PathBuf` ]] and [[ https:\/\/doc.rust-lang.org\/stable\/std\/path\/struct.Path.html | `&Path` ]] for paths and [[ https:\/\/doc.rust-lang.org\/stable\/std\/ffi\/struct.OsString.html | `OsString` ]] and [[ https:\/\/doc.rust-lang.org\/stable\/std\/ffi\/struct.OsStr.html | `&OsStr` ]] for strings (owned and references respectively. They do have os-specific extensions but as long as you don't use them it should be cross platform. That being said, if you are serializing and deserializing them you may need to write some platform dependant code.","author":"kevincox","id":"43991","dateCreated":"1520536377","dateModified":"1520536377"},{"type":"comment","comment":">>! In D2057#43989, @durin42 wrote:\n>>>! In D2057#43988, @kevincox wrote:\n>>>>! In D2057#43987, @durin42 wrote:\n>>> \n>>> Mercurial tries to be principled about always treating filenames as bytes. AIUI https:\/\/www.mercurial-scm.org\/wiki\/WindowsUTF8Plan is still the plan of record there?\n>> \n>> Reading that page it seems to claim that filenames should be utf8, not bytes. If utf8, this is what the code does, but if it is bytes that definitely won't work.\n> \n> IIRC it's bytes everyplace except Windows, where we pretend utf8 is real?\n> \n> We may have to make adjustments to this plan on macOS with APFS, but I'm not sure about that yet.\n\nI think we want to express a path as a dedicated type which has different underlying storage depending on the platform (bytes on Linux, UTF-16 on Windows). All filesystem operations should take a `Path` instance to operate on. This is the only way to cleanly round trip filenames between the OS, the application, and back to the OS. That leaves us with the hard problem of normalizing Mercurial's storage representation of paths (bytes) with the operating system's. But this world is strictly better than today, where we lose path data from the OS because we use POSIX APIs.\n\nFWIW, Python 3 rewrote the I\/O layer to use Win32 APIs everywhere. Combined with the `pathlib` types, I'm pretty sure Python 3 can round trip paths on Windows. I also think Rust's path type(s) have OS-dependent functionality.","author":"indygreg","id":"43990","dateCreated":"1520534077","dateModified":"1520534077"},{"type":"comment","comment":">>! In D2057#43988, @kevincox wrote:\n>>>! In D2057#43987, @durin42 wrote:\n>> \n>> Mercurial tries to be principled about always treating filenames as bytes. AIUI https:\/\/www.mercurial-scm.org\/wiki\/WindowsUTF8Plan is still the plan of record there?\n> \n> Reading that page it seems to claim that filenames should be utf8, not bytes. If utf8, this is what the code does, but if it is bytes that definitely won't work.\n\nIIRC it's bytes everyplace except Windows, where we pretend utf8 is real?\n\nWe may have to make adjustments to this plan on macOS with APFS, but I'm not sure about that yet.","author":"durin42","id":"43989","dateCreated":"1520533145","dateModified":"1520533145"},{"type":"comment","comment":">>! In D2057#43987, @durin42 wrote:\n> \n> Mercurial tries to be principled about always treating filenames as bytes. AIUI https:\/\/www.mercurial-scm.org\/wiki\/WindowsUTF8Plan is still the plan of record there?\n\nReading that page it seems to claim that filenames should be utf8, not bytes. If utf8, this is what the code does, but if it is bytes that definitely won't work.","author":"kevincox","id":"43988","dateCreated":"1520532520","dateModified":"1520532520"},{"type":"comment","comment":">>! In D2057#43892, @kevincox wrote:\n> On a higher level, all of these code appears to be treating file names as strings. This isn't really true and will disallow some valid file names. Maybe we should stick with bytes throughout. Of course this makes windows filepaths difficult because they are actually (utf16) strings.\n\nMercurial tries to be principled about always treating filenames as bytes. AIUI https:\/\/www.mercurial-scm.org\/wiki\/WindowsUTF8Plan is still the plan of record there?","author":"durin42","id":"43987","dateCreated":"1520531899","dateModified":"1520531899"},{"type":"inline","comment":"Take a `&str`.","replyTo":null,"isNewFile":"1","line":"62","lineLength":"0","path":"rust\/hgstorage\/src\/path_encoding.rs","diffId":"6724","author":"kevincox","id":"43985","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"```\nfn escape(out: &mut String, b: char) {\n    unimplemented!()\n}\n\npub fn encode_path(path: &str) -> String {\n    let mut out = String::with_capacity(path.len());\n\n    for c in path.bytes() {\n        let c = c as char;\n        match c {\n            'A'...'Z' => {\n                out.push('_');\n                out.push(c.to_ascii_lowercase());\n            }\n            '\\\\' | ':' | '*' | '?' | '\"' | '<' | '>' | '|' => {\n                escape(&mut out, c);\n            }\n            \/\/ The rest of the printable range.\n            ' '...'~' => {\n                out.push(c);\n            }\n            _ => {\n                escape(&mut out, c);\n            }\n        }\n    }\n\n    out\n}\n```\nhttps:\/\/godbolt.org\/g\/3WCQs3","replyTo":null,"isNewFile":"1","line":"57","lineLength":"0","path":"rust\/hgstorage\/src\/path_encoding.rs","diffId":"6724","author":"kevincox","id":"43984","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Don't pass a `char` by reference. Also it seems your function wants a `u8`.","replyTo":null,"isNewFile":"1","line":"11","lineLength":"0","path":"rust\/hgstorage\/src\/path_encoding.rs","diffId":"6724","author":"kevincox","id":"43983","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Use a String.","replyTo":null,"isNewFile":"1","line":"35","lineLength":"0","path":"rust\/hgstorage\/src\/path_encoding.rs","diffId":"6724","author":"kevincox","id":"43982","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"`encode_file_name`?","replyTo":null,"isNewFile":"1","line":"34","lineLength":"0","path":"rust\/hgstorage\/src\/path_encoding.rs","diffId":"6724","author":"kevincox","id":"43981","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Take a `&str`","replyTo":null,"isNewFile":"1","line":"34","lineLength":"0","path":"rust\/hgstorage\/src\/path_encoding.rs","diffId":"6724","author":"kevincox","id":"43980","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"```\np.ends_with(\".i\") || p.ends_with(\".d\")\n```","replyTo":null,"isNewFile":"1","line":"25","lineLength":"0","path":"rust\/hgstorage\/src\/path_encoding.rs","diffId":"6724","author":"kevincox","id":"43979","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"This isn't necessary.","replyTo":null,"isNewFile":"1","line":"23","lineLength":"0","path":"rust\/hgstorage\/src\/path_encoding.rs","diffId":"6724","author":"kevincox","id":"43978","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"I don't think you need this.","replyTo":null,"isNewFile":"1","line":"22","lineLength":"0","path":"rust\/hgstorage\/src\/path_encoding.rs","diffId":"6724","author":"kevincox","id":"43977","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"`Vec<char>` is odd. Is there any reason not to use a `String` or `Vec<u8>`","replyTo":null,"isNewFile":"1","line":"11","lineLength":"0","path":"rust\/hgstorage\/src\/path_encoding.rs","diffId":"6724","author":"kevincox","id":"43976","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"c should be a `u8`.","replyTo":null,"isNewFile":"1","line":"7","lineLength":"0","path":"rust\/hgstorage\/src\/path_encoding.rs","diffId":"6724","author":"kevincox","id":"43975","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"const HEX_DIGIT: [u8; 16] = *b\"0123456789abcdef\";","replyTo":null,"isNewFile":"1","line":"5","lineLength":"0","path":"rust\/hgstorage\/src\/path_encoding.rs","diffId":"6724","author":"kevincox","id":"43974","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"```\nassert!(!frags.is_empty());\n```","replyTo":null,"isNewFile":"1","line":"137","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"kevincox","id":"43973","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Please explain.","replyTo":null,"isNewFile":"1","line":"120","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"kevincox","id":"43972","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Make this one line and don't bother renaming.","replyTo":null,"isNewFile":"1","line":"86","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"kevincox","id":"43971","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"```\nfor &Fragment{frag_len, frag_ofs} in list.iter().rev()\n```","replyTo":null,"isNewFile":"1","line":"68","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"kevincox","id":"43970","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"It seems weird to take a cursor to a vec if you are just going to do an absolute seek. Can it work with `&mut [u8]`?","replyTo":null,"isNewFile":"1","line":"51","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"kevincox","id":"43969","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"`vec![0; count]` works. (The arguments might be the other way around).","replyTo":null,"isNewFile":"1","line":"54","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"kevincox","id":"43968","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"`mov` is overly shortened and generic.","replyTo":null,"isNewFile":"1","line":"51","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"kevincox","id":"43967","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"`pull` is very generic.","replyTo":null,"isNewFile":"1","line":"35","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"kevincox","id":"43966","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"```\nstruct Fragment {\n    len: u32,\n    offset: u32,\n}\n```","replyTo":null,"isNewFile":"1","line":"24","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"kevincox","id":"43965","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"s\/f\/fragment\/","replyTo":null,"isNewFile":"1","line":"40","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"kevincox","id":"43964","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"If you are unwrapping the `pop` there is no need for the prior check.","replyTo":null,"isNewFile":"1","line":"40","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"kevincox","id":"43963","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Maybe it's just me but I think it is more common to put the source before the destination.","replyTo":null,"isNewFile":"1","line":"35","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"kevincox","id":"43962","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"```\nassert!(!src.is_empty())\n```","replyTo":null,"isNewFile":"1","line":"39","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"kevincox","id":"43961","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Spell these out please.","replyTo":null,"isNewFile":"1","line":"13","lineLength":"0","path":"rust\/hgstorage\/src\/mpatch.rs","diffId":"6724","author":"kevincox","id":"43960","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"I would do `self.inner.map(|m| m.is_match(rp)).unwrap_or(false)` but this is fine.","replyTo":null,"isNewFile":"1","line":"200","lineLength":"0","path":"rust\/hgstorage\/src\/matcher.rs","diffId":"6724","author":"kevincox","id":"43959","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"s\/rp\/path\/","replyTo":null,"isNewFile":"1","line":"195","lineLength":"0","path":"rust\/hgstorage\/src\/matcher.rs","diffId":"6724","author":"kevincox","id":"43958","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"s\/relglob\/relative_glob_re\/","replyTo":null,"isNewFile":"1","line":"108","lineLength":"0","path":"rust\/hgstorage\/src\/matcher.rs","diffId":"6724","author":"kevincox","id":"43957","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Is this a warning or error? You might want to switch to `panic!`.","replyTo":null,"isNewFile":"1","line":"159","lineLength":"0","path":"rust\/hgstorage\/src\/matcher.rs","diffId":"6724","author":"kevincox","id":"43956","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"I would move this into the following match because it dedupes the `starts_with` check and puts the logic closer together.","replyTo":null,"isNewFile":"1","line":"160","lineLength":"0","path":"rust\/hgstorage\/src\/matcher.rs","diffId":"6724","author":"kevincox","id":"43955","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"s\/ln\/line\/","replyTo":null,"isNewFile":"1","line":"143","lineLength":"0","path":"rust\/hgstorage\/src\/matcher.rs","diffId":"6724","author":"kevincox","id":"43954","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Better name please.","replyTo":null,"isNewFile":"1","line":"131","lineLength":"0","path":"rust\/hgstorage\/src\/matcher.rs","diffId":"6724","author":"kevincox","id":"43953","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"You should be able to do `&string` rather then `string.as_str()` as it coerces. ","replyTo":null,"isNewFile":"1","line":"111","lineLength":"0","path":"rust\/hgstorage\/src\/matcher.rs","diffId":"6724","author":"kevincox","id":"43952","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"If you are just doing one call just return the result.","replyTo":null,"isNewFile":"1","line":"111","lineLength":"0","path":"rust\/hgstorage\/src\/matcher.rs","diffId":"6724","author":"kevincox","id":"43951","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"If you are going to call `String.as_str()` just take a `&str`.","replyTo":null,"isNewFile":"1","line":"108","lineLength":"0","path":"rust\/hgstorage\/src\/matcher.rs","diffId":"6724","author":"kevincox","id":"43950","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Can you manage a `&[u8]` rather then pointer arithmetic for the whole string. It will make me feel better and will probably be easier to read.","replyTo":null,"isNewFile":"1","line":"14","lineLength":"0","path":"rust\/hgstorage\/src\/matcher.rs","diffId":"6724","author":"kevincox","id":"43949","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Might be worth calling `String::with_capacity(pat.len())` since it will be at least that long.","replyTo":null,"isNewFile":"1","line":"10","lineLength":"0","path":"rust\/hgstorage\/src\/matcher.rs","diffId":"6724","author":"kevincox","id":"43948","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"s\/pat\/glob\/","replyTo":null,"isNewFile":"1","line":"9","lineLength":"0","path":"rust\/hgstorage\/src\/matcher.rs","diffId":"6724","author":"kevincox","id":"43947","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"What are these numbers?","replyTo":null,"isNewFile":"1","line":"49","lineLength":"0","path":"rust\/hgstorage\/src\/manifest.rs","diffId":"6724","author":"kevincox","id":"43946","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"s\/ent\/entry\/","replyTo":null,"isNewFile":"1","line":"42","lineLength":"0","path":"rust\/hgstorage\/src\/manifest.rs","diffId":"6724","author":"kevincox","id":"43945","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"What are these magic numbers?","replyTo":null,"isNewFile":"1","line":"33","lineLength":"0","path":"rust\/hgstorage\/src\/manifest.rs","diffId":"6724","author":"kevincox","id":"43944","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"This test has no assetions. Consider calling it `test_create_...` or something to indicate that you are just checking for panics.","replyTo":null,"isNewFile":"1","line":"155","lineLength":"0","path":"rust\/hgstorage\/src\/local_repo.rs","diffId":"6724","author":"kevincox","id":"43943","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Why does it need to be mutable to clone?","replyTo":null,"isNewFile":"1","line":"136","lineLength":"0","path":"rust\/hgstorage\/src\/local_repo.rs","diffId":"6724","author":"kevincox","id":"43942","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"s\/fp\/path\/","replyTo":null,"isNewFile":"1","line":"121","lineLength":"0","path":"rust\/hgstorage\/src\/local_repo.rs","diffId":"6724","author":"kevincox","id":"43941","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"`gd` is cryptic.","replyTo":null,"isNewFile":"1","line":"129","lineLength":"0","path":"rust\/hgstorage\/src\/local_repo.rs","diffId":"6724","author":"kevincox","id":"43940","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"```\nassert!(abspath.exists(), \"path not exists: {:?}\", abspath);\n```","replyTo":null,"isNewFile":"1","line":"127","lineLength":"0","path":"rust\/hgstorage\/src\/local_repo.rs","diffId":"6724","author":"kevincox","id":"43939","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"```\nwhile !root.join(\".hg\").exists() {\n    root = root.parent().expect(\".hg folder not found\");\n}\n```","replyTo":null,"isNewFile":"1","line":"66","lineLength":"0","path":"rust\/hgstorage\/src\/local_repo.rs","diffId":"6724","author":"kevincox","id":"43938","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"I would replace the condition with.\n\n```\nassert!(dot_hg_path.exists(), \".hg folder not found for the path given by -R argument: {:?}\", p);\n```","replyTo":null,"isNewFile":"1","line":"50","lineLength":"0","path":"rust\/hgstorage\/src\/local_repo.rs","diffId":"6724","author":"kevincox","id":"43937","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"You can add a later `.arg(dst)` to support non-utf8 paths instead of converting to a str here.","replyTo":null,"isNewFile":"1","line":"54","lineLength":"0","path":"rust\/hgstorage\/src\/lib.rs","diffId":"6724","author":"kevincox","id":"43936","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"In rust we generally avoid brackets around `as` as it is very tightly binding.","replyTo":null,"isNewFile":"1","line":"206","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43935","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Please use a better name for `sent`.","replyTo":null,"isNewFile":"1","line":"199","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43934","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"You can use the entry api here.","replyTo":null,"isNewFile":"1","line":"184","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43933","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"s\/rem\/path\/ or remaining_path.","replyTo":null,"isNewFile":"1","line":"183","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43932","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Use an `else if`.","replyTo":null,"isNewFile":"1","line":"175","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43931","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"You could do the following for a slight performance win and save a line.\n\n```\nif let Occupied(entry) = self.dmap.entry(relpath) {\n   ...\n}\n```","replyTo":null,"isNewFile":"1","line":"170","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43930","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"I recommend not renaming this. It is confusing.","replyTo":null,"isNewFile":"1","line":"5","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43929","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"This could have a better name.","replyTo":null,"isNewFile":"1","line":"48","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43928","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"```\nlet mut grey = Set::new();\ngrey.extend(self.dmap.keys().map(|s| s.as_path()));\n```\n\nAlso I would pick a name like `undiscovered_paths` or something. `grey` is cryptic.","replyTo":null,"isNewFile":"1","line":"146","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43927","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"I would also put this filter above. But more importantly all `_is_bad()` does is check for file types. So it seems like the former filter is redundant with this one.","replyTo":null,"isNewFile":"1","line":"169","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43926","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"I would move this filter beside the filter in the loop.","replyTo":null,"isNewFile":"1","line":"167","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43925","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"I would just call this `path` or `pathbuf`.","replyTo":null,"isNewFile":"1","line":"162","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43924","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Please explain why you are ignoring the error condition.","replyTo":null,"isNewFile":"1","line":"161","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43923","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"This is probably worth a helper function.","replyTo":null,"isNewFile":"1","line":"155","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43922","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"I would prefer doing the filter before the loop and storing it in a variable.","replyTo":null,"isNewFile":"1","line":"152","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43921","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"s\/mtc\/matcher\/","replyTo":null,"isNewFile":"1","line":"141","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43920","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Don't use `_` prefix for privates. Rely on rust viability. \n\nAlso `is_bad` isn't very informative.","replyTo":null,"isNewFile":"1","line":"130","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43919","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Is ignoring duplicate entries desired? It might be worth a comment explaining why.","replyTo":null,"isNewFile":"1","line":"125","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43918","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"This should Probably return a `Result<Self>` and pass the error to the caller.","replyTo":null,"isNewFile":"1","line":"87","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43917","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"1. Does this function need to be public? It seems internal to the constructor.\n1. If it doesn't need to be I would prefer it return the Map so that you don't have a partial-constructed DirState.","replyTo":null,"isNewFile":"1","line":"108","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43916","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"I would skip this check and rely on `p.metadata()`. Just switch `.unwrap()` to `.expect()` with a nicer message. This also handles race conditions more nicely.","replyTo":null,"isNewFile":"1","line":"90","lineLength":"0","path":"rust\/hgstorage\/src\/dirstate.rs","diffId":"6724","author":"kevincox","id":"43915","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Is this used yet? It probably also needs some documentation because I don't really understand the fields (but I do have little domain knowledge).","replyTo":null,"isNewFile":"1","line":"49","lineLength":"0","path":"rust\/hgstorage\/src\/config.rs","diffId":"6724","author":"kevincox","id":"43914","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"If you are just going to convert to String I would recommend taking a String argument.\n\nAlso prefer `.to_owned()` over `.to_string()`.","replyTo":null,"isNewFile":"1","line":"78","lineLength":"0","path":"rust\/hgstorage\/src\/config.rs","diffId":"6724","author":"kevincox","id":"43913","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Just put `msg: content` in the struct construction.","replyTo":null,"isNewFile":"1","line":"33","lineLength":"0","path":"rust\/hgstorage\/src\/changelog.rs","diffId":"6724","author":"kevincox","id":"43912","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"If you aren't using the value I would prefer `truncate(NodeId::hex_len())`","replyTo":null,"isNewFile":"1","line":"31","lineLength":"0","path":"rust\/hgstorage\/src\/changelog.rs","diffId":"6724","author":"kevincox","id":"43911","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Passing a message as a third argument is really useful.","replyTo":null,"isNewFile":"1","line":"24","lineLength":"0","path":"rust\/hgstorage\/src\/changelog.rs","diffId":"6724","author":"kevincox","id":"43910","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"Let rust do the overflow checking.\n\n```\nacc = acc.checked_mul(85)\n    .ok_or_else(|| {\n        PyErr::new::<exc::ValueError, _>(\n            py,\n            format!(\"bad base85 character at position {}\", i))\n    })?;","replyTo":null,"isNewFile":"1","line":"152","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43909","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"This is probably worth a comment that this is safe because D85DEC is required to be initialized before this function is called.","replyTo":null,"isNewFile":"1","line":"91","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43908","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"```while !ptext.is_empty()```","replyTo":null,"isNewFile":"1","line":"52","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43907","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"It might be best to use a [[ https:\/\/doc.rust-lang.org\/stable\/std\/io\/struct.Cursor.html | `std::io::Cursor` ]] and let it drack `dst_off` for your.","replyTo":null,"isNewFile":"1","line":"47","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43906","dateCreated":"1520530210","dateModified":"1520530210"},{"type":"inline","comment":"`ptext` isn't very descriptive. ","replyTo":null,"isNewFile":"1","line":"45","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43905","dateCreated":"1520530209","dateModified":"1520530209"},{"type":"inline","comment":"Tracking len manually is a smell. Why not drop it and use `ptest.is_empty()`.","replyTo":null,"isNewFile":"1","line":"63","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43904","dateCreated":"1520530209","dateModified":"1520530209"},{"type":"inline","comment":"I would prefer the name `chunk` or even `accum` is a lot mode obvious to me than `acc`.","replyTo":null,"isNewFile":"1","line":"54","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43903","dateCreated":"1520530209","dateModified":"1520530209"},{"type":"inline","comment":"I suspect this type annotation isn't required.","replyTo":null,"isNewFile":"1","line":"47","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43902","dateCreated":"1520530209","dateModified":"1520530209"},{"type":"inline","comment":"Why the braces here?","replyTo":null,"isNewFile":"1","line":"46","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43901","dateCreated":"1520530209","dateModified":"1520530209"},{"type":"inline","comment":"IIUC pad is only ever checked `== 0`. Can it be made into a bool?","replyTo":null,"isNewFile":"1","line":"23","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43900","dateCreated":"1520530209","dateModified":"1520530209"},{"type":"inline","comment":"I would just combine these into one line as the name `ch` isn't adding much.","replyTo":null,"isNewFile":"1","line":"58","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43899","dateCreated":"1520530209","dateModified":"1520530209"},{"type":"inline","comment":"`&str` can only hold valid utf8 data? Does it make more sense to have `&[u8]` here for a list of bytes?","replyTo":null,"isNewFile":"1","line":"23","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43898","dateCreated":"1520530209","dateModified":"1520530209"},{"type":"inline","comment":"Would it be possible to separate the decode from the python objects. I'm thinking two helper functions.\n\n```\nfn b85_required_len(text: &str) -> usize\nfn b85_encode(text: &str, pad: i32, out: &mut [u8]) -> Result<()>\n```\n\n","replyTo":null,"isNewFile":"1","line":"23","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43897","dateCreated":"1520530209","dateModified":"1520530209"},{"type":"inline","comment":"```\nfor i in &[24, 16, 8, 0]\n```","replyTo":null,"isNewFile":"1","line":"56","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43896","dateCreated":"1520530209","dateModified":"1520530209"},{"type":"inline","comment":"I prefer something like this: https:\/\/play.rust-lang.org\/?gist=5ca18a5b95335600e911b8f9310ea5c7&version=stable\n\nI doubt lazy_static is too slow. Otherwise we can stay like this until const functions get implemented.\n\nEither way note:\n* I changed the type of B85CHARS to an array as opposed to an array ref.\n* The loop condition is much nicer.","replyTo":null,"isNewFile":"1","line":"21","lineLength":"0","path":"rust\/hgbase85\/src\/base85.rs","diffId":"6724","author":"kevincox","id":"43895","dateCreated":"1520530209","dateModified":"1520530209"},{"type":"inline","comment":"If this is going to be reused I would move it into it's own crate. It seems like everything here could be boiled down to a single function call in main.","replyTo":"43863","isNewFile":"1","line":"1","lineLength":"0","path":"rust\/hgbase85\/build.rs","diffId":"6724","author":"kevincox","id":"43894","dateCreated":"1520530209","dateModified":"1520530209"},{"type":"comment","comment":"I will try to finish the review later, but it might be quicker if you incorporate some of the changes first since a lot of them are repeated many times. Overall it looks good, there are a couple of things that i would highlight to make the code easier to read.\n\n* Prefer more descriptive variable names.\n* If you can, avoid \"pointer\" arithmetic. Cursors and slices are nice and have great convenience methods.\n* Group your flow control and filtering more.\n* Try to keep your types straight. In rust using the right types helps catch errors. So be aware of `char` vs `u8` vs `String` vs `Vec<char>` vs `Vec<u8>`.\n\nOn a higher level, all of these code appears to be treating file names as strings. This isn't really true and will disallow some valid file names. Maybe we should stick with bytes throughout. Of course this makes windows filepaths difficult because they are actually (utf16) strings.","author":"kevincox","id":"43893","dateCreated":"1520530209","dateModified":"1520530209"},{"type":"reject","author":"kevincox","id":"43892","dateCreated":"1520530209","dateModified":"1520530209"},{"type":"comment","comment":"Doesn't mononoke have code to read revlogs already?","author":"glandium","id":"43869","dateCreated":"1520495348","dateModified":"1520495348"},{"type":"inline","comment":"IIRC, core Mercurial keeps an open file handle on revlogs and ensures we don't run out of file handles by not keeping too many revlogs open at the same time. For scanning operations, not having to open and close the file handles all the time will make a difference for performance. Also, core Mercurial loads the entirety of the `.i` file into memory. That's a scaling problem for large revlogs. But it does make performance of index lookups really fast.","replyTo":null,"isNewFile":"1","line":"279","lineLength":"0","path":"rust\/hgstorage\/src\/revlog_v1.rs","diffId":"6724","author":"indygreg","id":"43867","dateCreated":"1520494272","dateModified":"1520494272"},{"type":"inline","comment":"A thread pool to help with zlib decompression should go a long way here.\n\nProbably too early to think about this, but we'll likely eventually want a global thread pool for doing I\/O and CPU expensive tasks, such as reading chunks from a revlog and decompressing them.\n\nFWIW, we're going to radically alter the storage format in order to better support shallow clones. But that work hasn't started yet. I still think there is a benefit to implementing the revlog code in Rust though.","replyTo":null,"isNewFile":"1","line":"290","lineLength":"3","path":"rust\/hgstorage\/src\/revlog_v1.rs","diffId":"6724","author":"indygreg","id":"43866","dateCreated":"1520494272","dateModified":"1520494272"},{"type":"inline","comment":"I would not worry about supporting v0 or v2 at this time. v0 is only important for backwards compatibility with ancient repos. And v2 never got off the ground.","replyTo":null,"isNewFile":"1","line":"95","lineLength":"0","path":"rust\/hgstorage\/src\/config.rs","diffId":"6724","author":"indygreg","id":"43865","dateCreated":"1520494272","dateModified":"1520494272"},{"type":"inline","comment":"This is definitely nifty and an impressive achievement \\o\/\n\nThe `r-` commands for testing pure Rust code paths are an interesting idea!\n\nI think I'm OK with including support for this in `hgcli`. But I think the code should live in a separate file so it doesn't pollute `main()`. And it should be behind a Cargo feature flag so we maintain compatibility with `hg` as much as possible by default.\n\nAlso, Mercurial's command line parser is extremely wonky and has some questionable behavior. If the intent is to make `rhg` compatible with `hg`, we would need to preserve this horrible behavior. We'll likely have to write a custom argument parser because of how quirky Mercurial's argument parser is :(","replyTo":null,"isNewFile":"1","line":"233","lineLength":"28","path":"rust\/hgcli\/src\/main.rs","diffId":"6724","author":"indygreg","id":"43864","dateCreated":"1520494272","dateModified":"1520494272"},{"type":"inline","comment":"I see this file was copied. There's nothing wrong with that. But does this mean we will need a custom build.rs for each Rust package doing Python? If that's the case, then I would prefer to isolate all our rust-cpython code to a single package, if possible. I'm guessing that could be challenging due to crossing create boundaries. I'm sure there are placed where we don't want to expose symbols outside the crate.\n\nI'm curious how others feel about this.","replyTo":null,"isNewFile":"1","line":"1","lineLength":"0","path":"rust\/hgbase85\/build.rs","diffId":"6724","author":"indygreg","id":"43863","dateCreated":"1520494272","dateModified":"1520494272"},{"type":"comment","comment":"First of all - wow! Thanks for writing all this code. There's definitely a lot to work with. And work with it we will!\n\nThis is definitely too big to review as one commit. If you could do *any* work to split it up, it would be greatly appreciated. I'd focus on the pure Rust pieces first. Everything needed to open revlogs would be great!\n\nYou may find our custom Phabricator extensions (linked from https:\/\/www.mercurial-scm.org\/wiki\/Phabricator) useful for submitting series of commits to Phabricator.\n\nRegarding the performance, that's pretty good! The dirstate code is some of the most optimized code in Mercurial. There are some gnarly Python C hacks to make it fast. Some of those tricks involve using special system calls to walk directories to minimize the number of system calls. I'm not sure if the crate you imported has those optimizations. (I wouldn't be surprised either way.) I wouldn't worry too much about performance at this juncture. But I suspect we could make the Rust code another 50% faster with some tweaking. It would also be interesting to test on a larger repo, say https:\/\/hg.mozilla.org\/mozilla-unified. Also, I believe there are hooks in the dirstate code to use Watchman (fsmonitor). Those hooks are critical in order to achieve peak performance on large repositories.\n\nSince you seem to be proficient at writing lots of Rust code, if you are looking for another project, may I suggest porting `chg` to Rust? That code is in `contrib\/chg`. That might be the easiest component to actually ship in Rust since it is a standalone binary that doesn't link against Python. But we shouldn't get ahead of ourselves :)\n\nAnyway, it is late for me and I need to detach from my computer. I'm sure others will have things to say as well...","author":"indygreg","id":"43862","dateCreated":"1520494272","dateModified":"1520494272"},{"type":"comment","comment":"Hi all,\n\nBased on the discussion a few weeks ago, I come up with a solution for review and discussion. After reading the Oxidation plan, the first thought is to bypass python engine and current plugin system IFF on request. If user (maybe background checker of IDE) request r-* subcommands, then hg gives rust implementations instead of python's. So I try to make hg r-status as fast as possible. The submitted version has comparable performance (as an example of the performance, not evidence, on my MacBook, in hg's own repo, hg r-status 150ms, and hg status 220ms). I am using CodeXL to profile the performance, and plan to use Future.rs to make the loading parallel and maybe 30ms faster.\n\nThe implementation originates from hg python implementation, because the python version is really fast. I tried to split into small changes, however, I eventually to combine all hgstorage module as one commit.\n\nThank you for your comments!","author":"Ivzhh","id":"43861","dateCreated":"1520491361","dateModified":"1520491361"},{"type":"comment","comment":"  - merge with stable\n  - translate base85.c into rust code\n  - move hgbase85 into independent module\n  - add hgstorage crate\n  - hg status implementation in rust\n","author":"Ivzhh","id":"43856","dateCreated":"1520490633","dateModified":"1520490633"},{"type":"update","diffId":"6724","author":"Ivzhh","id":"43855","dateCreated":"1520490633","dateModified":"1520490633"},{"type":"comment","comment":"Thank you @indygreg!\n\nThe OxidationPlan is my best reference when I started to make a move, and this thread is even more helpful. I am really interested in exploring this ;-) In 2014 I was trying to change the hg backend storage to Postgres, a silly and failed experiment.\n\nAnyway, I will save everyone's time and stop talking. I will come back later with a more meaningful implementation.","author":"Ivzhh","id":"34977","dateCreated":"1518075333","dateModified":"1518075333"},{"type":"comment","comment":"To be honest, we're not yet sure what we'll decide for the Python -> Rust bridge. The problem is summarized in the `Rust <=> Python Interop` section on https:\/\/www.mercurial-scm.org\/wiki\/OxidationPlan.\n\nI suspect at some level we'll need a CPython extension for CPython for performance reasons (especially for high volume function calls). PyPy obviously uses CFFI. I think the ideal outcome is we can write Rust that exposes a C API and use CFFI natively on PyPy and something like `cbindgen` + `Milksnake` to auto-generate a CPython extension that acts as a wrapper around the C API exposed by Rust. I'm not sure if anyone has invented this exact wheel yet. If not, it's probably faster to use `rust-cpython`. Maybe several months from now we have enough Rust and maintaining `rust-cpython` is painful enough that we pursue the auto-generated CPython extension route.\n\nWhat I'm trying to say is you have a green field to explore! But at this juncture, perfect is the enemy of done. We'll be happy with any forward progress, even failed experiments.","author":"indygreg","id":"34976","dateCreated":"1518074550","dateModified":"1518074550"},{"type":"comment","comment":"As the author of this patch, actually I have the same concern. I started to translate base85 as baby steps to find a way of integrating rust and cpython, on my side, Today I modify setup.py, policy.py and makefile to run hg's test suit with the new base85. For myself, it is only proof of concept.\n\nMaybe I should take another way: translate more python modules into CFFI-style, and let CFFI call rust implementation. And gradually change more implementations of python modules with corresponding cffi-style, while keep the python interface the same. My own hope is the rust routines will be able to call each other and eventually run some __basic__ tasks without calling python part. And the rust still lazily provides info to python interface for extensions etc.\n\nI am exploring this way now, and hope the findings will be useful for community to make decision.\n\nThank you all for the comments!","author":"Ivzhh","id":"34975","dateCreated":"1518070365","dateModified":"1518070365"},{"type":"comment","comment":"What would be the advantage of taking this? Since we already have the C implementation, it's not likely to gain us any performance. On the other hand, it might make a good test case for integrating Rust and Python, finding the right API boundaries and experimenting with different approaches, precisely \/\/because\/\/ we already have a C implementation. @indygreg @durin42 what are your thoughts about it?\n\n","author":"krbullock","id":"34910","dateCreated":"1518062354","dateModified":"1518062354"},{"type":"comment","comment":"I agree with the splitting comments :) In fact there might already be a base85 crate which can be used: https:\/\/docs.rs\/zero85. Either way I'll hold off on the review, feel free to ping me when you are ready for me to take a look.","author":"kevincox","id":"34640","dateCreated":"1518001891","dateModified":"1518001891"},{"type":"comment","comment":"Thank you @indygreg for your detailed explanation!\n\nI understand the process now, and I will go back reading the developer's guide thoroughly again. I will try my best to provide a relatively clean stack of patches.\n\nThank you for you time!","author":"Ivzhh","id":"34561","dateCreated":"1517972031","dateModified":"1517972031"},{"type":"comment","comment":"We generally prefer that patches to Mercurial be small and do a single thing. This makes it easier to review and understand changes, since each change can be evaluated in isolation. If you submit changesets together using `hg phabsend`, they automatically show up as a \/\/stack\/\/ in Phabricator. And if changesets at the bottom of the stack are ready to land, we generally land those without waiting for the entire stack to land. This enables forward progress to be made and this is generally better for everyone than waiting until a series of commits is perfect before adding any of them.\n\nWhat that means is you should ideally split this work into smaller parts. For example:\n\n  # Add the pure Rust code\/crate\n  # Add the Python Rust code\/crate\n  # Build system \/ module policy changes\n\nI'm not sure of the order of things though. Since this is the first Rust extension, it's not clear what needs to be implemented in what order. I'm fine looking at a large commit if things are too tightly coupled to separate. But you should strive to make smaller commits.\n\n","author":"indygreg","id":"34557","dateCreated":"1517970792","dateModified":"1517970792"},{"type":"comment","comment":"Sure, thank you for the comments! I can definitely prepare makefile and setup.py to make the building process work with rust part. I am planning to change the policy.py module to support and try to load rust modules and run all the tests. I will submit a new patch after finishing these two tasks.\n\nAfter reading wiki\/OxidationPlan again, I plan to change to cffi for better compatibility (pypy and others), and try to build algorithms in pure rust. Shall I wait till migrating to cffi based solution now and resubmit this patch with all three changes (building, testing, and cffi)?\n\nThank you!","author":"Ivzhh","id":"34556","dateCreated":"1517970271","dateModified":"1517970271"},{"type":"comment","comment":"Yes, we should definitely split things into multiple crates. Small, narrowly-focused crates does seem to be the Rust way, after all.\n\n`hgcli` should be for things specific to the Rust implementation of `hg`. I think this can also include the feature set of `chg` (once we've ported `chg` to Rust).\n\nI definitely support separating the \"pure Rust\" from the \"Python Rust\" via a crate boundary. It is generally useful to have Rust that isn't bound to Python because it will facilitate reuse outside of Python contexts. For example, someone could implement a Mercurial wire protocol server in pure Rust without needing to worry about Python. Of course, we're likely to encounter areas where we really want tight coupling in order to achieve optimal performance in Python. So we may have to design APIs on the pure Rust side to facilitate CPython use. I'm OK with that.\n\nAs for how many crates to have, I don't have super strong opinions. I could see us putting every little component\/subsystem in its own crate. I could also see us putting everything in one large crate. I don't think it is worth deciding at this early juncture. API design and ability to be reused outside its originally intended purpose is the important property to strive for. I think that has more to do with how the code is authored rather than which crates things are in.\n\nA missing piece of this patch is the build system and module loader integration. We have a \/\/module policy\/\/ that dictates which implementation of a Python module we use. We probably want to introduce a `rust` policy that uses Rust-based modules where available and falls back to the `cext` modules\/policy if a Rust module isn't available. We also need to figure out how to integrate Rust into `setup.py`. But I think the build system bit can be deferred until we're actually ready to ship Rust, which is still a bit of ways off. I'm happy for the workflow to be \/\/run cargo in order to load Rust modules\/\/ for the time being. But if you can implement `Makefile` and\/or `setup.py` integration to build these Rust extensions, that would be awesome.","author":"indygreg","id":"34484","dateCreated":"1517947275","dateModified":"1517947275"},{"type":"comment","comment":"I am open to the three-crates plan. Oirginally I have hgcli and hgext separately, and I was planning to use CFFI mode. I am a pypy user too, so I will be willing to provide a python C API free crate for pypy and others.","author":"Ivzhh","id":"34443","dateCreated":"1517928889","dateModified":"1517944524"},{"type":"inline","comment":"I think I'd like to separate things a bit more and have a Python-free module, and then a glue module that we can use to call into the pure Rust. Part of the reason is that in my perfect world we won't use the cpython crate for speedups so they can be used from pypy as well. Separating them at least makes it easier to have an extern \"C\" version of the method that can be used from cffi instead of only through the CPython API.\n\n(Not sure what opinions others have. It's likely that I'll attempt this approach in the near future as part of a continued attempt to speed up `hg diff`.)","replyTo":null,"isNewFile":"1","line":"22","lineLength":"0","path":"rust\/hgcli\/src\/hgext\/base85.rs","diffId":"5238","author":"durin42","id":"34390","dateCreated":"1517913293","dateModified":"1517913293"},{"type":"comment","comment":"I'd be curious to see what @indygreg has to say about this, maybe wait on his input before doing any work in response to my feedback?\n\nI do wonder if we should have at least three crates:\n\n # hgcli\n # libmercurial\n # hgcext\n\nThe first one would be the command-line entry point, the last could use the cpython API, and libmercurial would be \"pure rust\" and open the door to eventually having a libhg or something that exports C functions and would be suitable for cffi and linking into other binaries?","author":"durin42","id":"34389","dateCreated":"1517913293","dateModified":"1517913293"},{"type":"update","diffId":"5238","author":"Ivzhh","id":"34367","dateCreated":"1517869861","dateModified":"1517869861"}],"dateCreated":"1517869861","dateModified":"1521888284","status":"Needs Review"},{"id":"2884","callsign":"HG","title":"wireproto: experimental command to emit file data","author":"indygreg","summary":"Partial clones will require new wire protocol functionality to\nretrieve repository data. The remotefilelog extensions - which\nimplements various aspects of partial clone - adds a handful of\nwire protocol commands:\n\ngetflogheads\n   Obtain heads of a filelog\ngetfile\n   Obtain data for an individual file revision\ngetfiles\n   Batch version of getfile\ngetpackv1\n   Obtain a \"pack file\" containing index and data on multiple\n   files\n\n(among others)\n\nRecently, the wire protocol has gained support for \"obtain repository\ndata\" in the form of overloading the \"getbundle\" wire protocol\ncommand. This is arguaby OK in the context of \"all data is attached\nto bundles\" and \"bundles are a self-contained representation of\ncomplete repository data.\" But partial clone invalidates these\nassumptions because in a partial clone world, we no longer can assume\nthings like \"the client has all the base revisions.\"\n\nIn a partial clone world, we'll need wire protocol commands that allow\nclients to obtain specific pieces of data with vastly different\naccess patterns. For example, a client may want to obtain \"index\"\ndata but keep the fulltext data on the server. Or vice-versa. Or a\nclient may wish to fetch all revisions of a specific file but only\nthe latest revision of another. These access patterns will be\ndifficult to shoehorn into single, powerful commands (like\n\"getbundle\"). Even if we could, doing that isn't wise from a server\nimplementation perspective because it makes implementing scalable\nservers hard. We want server-side commands to be small and simple\nso alternate server implementations can come into existence more\neasily.\n\nThis is one reason why the frame-based wire protocol I'm implementing\nsupports command pipelining and out-of-order responses. This\nproperty will enable clients performing complex operations to send\ncommand streams containing dozens or even hundreds of small command\nrequests to servers.\n\nAnyway, this commit implements an experimental wire protocol command\nfor \"get files data.\" Essentially, you give it a changeset revision\nyou are interested in and it spits back all the files and their data\nin that revision, as fulltexts.\n\nThis command is just one way a server could emit data for files.\nA variation of this command that accepts specific file paths and nodes\nwhose data is to be retrieved would also be useful. And I imagine we'll\neventually implement that. It would also be useful to emit index\ndata. Or have each file blob be individually compressed. (Right now\ncompression is performed on the whole stream because that's how the\nwire protocol currently works - but I have plans to evolve the frame\nbased protocol to do new and novel things here.)\n\nI'm not even sure this variation of the wire protocol command is a\ngood one to have! One reason I want to start with this command is\nthat it seems like a useful primitive. For example, with this\ncommand, one could build a client that is able to realize a working\ndirectory from a single wire protocol request: you can literally\nstream the response to this command and turn the data into files on\nthe filesystem with minimal stream processing!\n\nAs implemented, this command is effectively a benchmark of revlog\nreading and\/or compression. On the mozilla-unified repository when\noperating on revision c488b8d0e074efb490ebca32db68eb77871bfd2f (a\nrecent revision of mozilla-central, the head of Firefox development),\nmy i7-6700K yields the following:\n\n* no compression: 1478MB;  ~94s wall; ~56s CPU\n* zstd level 3:    343MB;  ~97s wall; ~57s CPU\n* zlib level 6:    367MB; ~116s wall; ~74s CPU\n\nFor comparison, `hg bundle --base null -r c488b8d0e0 -t zstd-v2`\n(which approximates what `hg clone -r` would be doing on the server)\nyields:\n\n  1397MB; ~624s wall; ~225s CPU\n\nOf course, these are vastly different operations. But this does\ndemonstrate that if your use case of version control is \"check out\nrevision X\" and you were previously relying on `hg clone` [without\nstream clone bundles] to do that, this wire protocol command\nis overall much more efficient on servers. It's worth noting that\nthe use case of version control for many automated systems *is*\n\"check out revision X.\" So I think providing a clone mode that can\nrealize a working copy as fast as possible is a worthwhile feature\nto have!","testPlan":"","lineCount":"335","dependsOn":[],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"comment","comment":"This needs rebasing. Please hold off reviewing.","author":"indygreg","id":"47297","dateCreated":"1521732096","dateModified":"1521732096"},{"type":"plan-changes","author":"indygreg","id":"47296","dateCreated":"1521732096","dateModified":"1521732096"},{"type":"update","diffId":"7150","author":"indygreg","id":"46609","dateCreated":"1521503993","dateModified":"1521503993"},{"type":"update","diffId":"7081","author":"indygreg","id":"46289","dateCreated":"1521241628","dateModified":"1521241628"}],"dateCreated":"1521241628","dateModified":"1521732096","status":"Changes Planned"},{"id":"2588","callsign":"HG","title":"commit: adds multiline commit message support(issue5616)","author":"sangeet259","summary":"The earlier functionality used to ignore the all but last -m flag's value.\nIt used to store the arguments information in a state dict and then overwrites each\nsubsequent value of -m in the 'message' key. This patch intercepts the message flag and\nchecks if it is already empty. In case it is not empty, add the current value of -m flag to\nthe message key with a leading '\\n' character. This makes every subsequent passed -m flag\nas the new line message of the commit.","testPlan":"","lineCount":"28","dependsOn":[],"reviewers":[],"ccs":["durin42","tom.prince","yuja","pulkit","jeffpc","mercurial-devel"],"actions":[{"type":"comment","comment":">>! In D2588#44645, @sangeet259 wrote:\n> Since the current code just overwrites message each time with the newer. What can be done to avoid losing the earlier message values?\n\nYou should look how we handle multiple `--rev` flags. That will help you in this case.","author":"pulkit","id":"44677","dateCreated":"1520666278","dateModified":"1520666278"},{"type":"comment","comment":"Since the current code just overwrites message each time with the newer. What can be done to avoid losing the earlier message values?\n","author":"sangeet259","id":"44645","dateCreated":"1520655918","dateModified":"1520655918"},{"type":"inline","comment":"agreed","replyTo":"42705","isNewFile":"1","line":"847","lineLength":"0","path":"tests\/test-commit.t","diffId":"6537","author":"durin42","id":"43618","dateCreated":"1520437801","dateModified":"1520437801"},{"type":"comment","comment":"@pulkit Yeah, this rather a not so elegant hack. How can it be parsed in hg commit? \nAs the first message passed is overwritten by the second one at this step only.  :\/\n\n@yuja I think maybe subclassing the customopt may be a fair choice.","author":"sangeet259","id":"43614","dateCreated":"1520437520","dateModified":"1520437520"},{"type":"inline","comment":"Right. Alternatively, we could add a subclass of customopt, but I don't know which\nwould be nicer.\n\nFWIW, I'm kinda -1 on this feature, but I have no strong opinion and Git is the\ncurrent standard.","replyTo":"42704","isNewFile":"1","line":"365","lineLength":"0","path":"mercurial\/fancyopts.py","diffId":"6537","author":"yuja","id":"42772","dateCreated":"1520184244","dateModified":"1520184244"},{"type":"inline","comment":"starting with a newline seems awkward.","replyTo":null,"isNewFile":"1","line":"847","lineLength":"0","path":"tests\/test-commit.t","diffId":"6537","author":"pulkit","id":"42705","dateCreated":"1520181862","dateModified":"1520181862"},{"type":"inline","comment":"I think this is not the right place to have this hack. We should have this as a part of `hg commit` code rather.","replyTo":null,"isNewFile":"1","line":"365","lineLength":"0","path":"mercurial\/fancyopts.py","diffId":"6537","author":"pulkit","id":"42704","dateCreated":"1520181862","dateModified":"1520181862"},{"type":"update","diffId":"6537","author":"sangeet259","id":"42559","dateCreated":"1520149348","dateModified":"1520149348"},{"type":"comment","comment":"@jeffpc That means two new lines. Shall I edit this patch to effect that?\n","author":"sangeet259","id":"42558","dateCreated":"1520142981","dateModified":"1520147672"},{"type":"update","diffId":"6536","author":"sangeet259","id":"42557","dateCreated":"1520139879","dateModified":"1520139879"},{"type":"comment","author":"sangeet259","id":"41893","dateCreated":"1520098879","dateModified":"1520098898"},{"type":"comment","comment":"FWIW, this is slightly different behavior from what git does.  There, each -m is added as a separate *paragraph*.  IOW,\n\n\n```\n$ git commit -m \"first\" -m \"second\"\n```\n\ncreates:\n\n```\nfirst\n\nsecond\n```","author":"jeffpc","id":"41892","dateCreated":"1520098727","dateModified":"1520098727"},{"type":"update","diffId":"6438","author":"sangeet259","id":"41882","dateCreated":"1520098029","dateModified":"1520098029"}],"dateCreated":"1520098029","dateModified":"1520666278","status":"Needs Review"},{"id":"2681","callsign":"HG","title":"[PoC] scmutil: support local only obsolescence","author":"indygreg","summary":"Now that we have a config option for enabling local-only obsolescence,\nit is time to do something with it.\n\nThis commit teaches cleanupnodes() - which is called after rewrite\noperations - to handle local only obsolescence mode.\n\nIn this mode, we create a backup bundle of the obsoleted changesets -\njust like what happens if obsolescence is disabled. But we don't strip\nthe repo: we keep the original changesets around in a non-visible state.\n\nThe new code verifies that no unstable changesets are introduced in\nlocal-only obsolescence mode.\n\nThe new code hackily only runs if the action is \"amend.\" The intent\nis to make this conditional only on the feature option. However,\ndoing that would have significant test fallout. So we limit to \"amend\"\nfor now.\n\nTODO:\n\n* Better test coverage (I think the \"(testcase !)\" syntax might be\n  subtly wrong by flagging output as optional and not required).\n* Delete obsolescence markers when we pull and unbundle from the\n  bundle.\n* Support pulling locally-hidden heads.","testPlan":"","lineCount":"143","dependsOn":["2680"],"reviewers":[],"ccs":["mharbison72","mercurial-devel"],"actions":[{"type":"comment","comment":"re: the \"(testcase !)\" syntax- I agree that it would be better if \"(false !)\" means the line is *not* present, instead of optional.\n\nI made an attempt to do that, but somehow I came up with a test in test-run-tests.t that failed there, but the exact same test ran fine in another *.t.  I wondered if it was a case of run-tests.py processing it for a test inside test-run-tests.t, and then the main instance of run-tests.py re-processing it.  If I can find that code, I can post it if you think it might be useful.  But I've long forgotten the nuance in that code. ","author":"mharbison72","id":"43337","dateCreated":"1520216552","dateModified":"1520216552"},{"type":"update","diffId":"6639","author":"indygreg","id":"43226","dateCreated":"1520210577","dateModified":"1520210577"}],"dateCreated":"1520210577","dateModified":"1520216552","status":"Needs Review"},{"id":"2682","callsign":"HG","title":"[PoC] changegroup: delete obs markers when applying changegroup","author":"indygreg","summary":"In local-only obsolescence mode, our short-term hack to unhide a\nchangeset is to delete obsolescence markers causing it to be hidden.\nWe need to do this every time a changeset is introduced into the\nrepository.\n\nI was hoping we could implement the unhide logic once, around\ntransaction close time. However, it appears that we'll need to\nteach every mechanism that introduces changesets to unhide\nchangesets because of interaction between those mechanisms and\nvisibility. For example, changegroup application no-ops on incoming\nchangesets that are present but hidden. The code that scans for\n\"new\" changesets is based on changesets between the old and new\nchangelog length. Since the unhidden changeset is possibly older\nthan the old changelog length, we need to either track unhides\nexplicitly or deal with them specially.\n\nThis commit teaches changegroup application to record which changesets\nare in the incoming changegroup so it can scan for relevant\nobsolescence markers after changegroup applications and delete\ntheir markers.\n\nI'm not convinced this is the best approach to the problem. But it's\na start.","testPlan":"","lineCount":"27","dependsOn":["2681"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"6640","author":"indygreg","id":"43236","dateCreated":"1520210580","dateModified":"1520210580"}],"dateCreated":"1520210580","dateModified":"1520210580","status":"Needs Review"},{"id":"2680","callsign":"HG","title":"[PoC] obsolete: make markers database writable if local-only mode enabled","author":"indygreg","summary":" ","testPlan":"","lineCount":"4","dependsOn":["2679"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"6638","author":"indygreg","id":"43217","dateCreated":"1520210574","dateModified":"1520210574"}],"dateCreated":"1520210574","dateModified":"1520210577","status":"Needs Review"},{"id":"2622","callsign":"HG","title":"ui: adding a generic method to read config items with an arbitrary type","author":"rdamazio","summary":"This can be used for reading many config options of various types where the\ntype of each is known in some other structure (e.g. options matching command\nflags).","testPlan":"","lineCount":"10","dependsOn":[],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"6520","author":"rdamazio","id":"42415","dateCreated":"1520121279","dateModified":"1520121279"}],"dateCreated":"1520121279","dateModified":"1520121282","status":"Needs Review"},{"id":"2079","callsign":"HG","title":"color: honor NO_COLOR","author":"indygreg","summary":"The http:\/\/no-color.org\/ initiative is trying to get programs that\nemit color by default to honor a NO_COLOR environment variable to\ndisable color.\n\nI think that's a good idea. So this commit implements support for\nNO_COLOR.\n\nI'm not sure if the precedence of settings is proper here. Right now,\nNO_COLOR overrides config settings set by hgrc or --config. But it\ndoesn't override --color. I can see an argument for honoring\n--config as well. Same for hgrc (since color is enabled by default\nthese days). But the existing logic\/precedence is unclear to me. So\nI went with an easy implementation.","testPlan":"","lineCount":"43","dependsOn":[],"reviewers":["lothiraldan","yuja"],"ccs":["durin42","martinvonz","yuja","quark","lothiraldan","mercurial-devel"],"actions":[{"type":"comment","comment":"(just clarify the current state of this patch)","author":"yuja","id":"37633","dateCreated":"1518697359","dateModified":"1518697359"},{"type":"reject","author":"yuja","id":"37632","dateCreated":"1518697359","dateModified":"1518697359"},{"type":"comment","comment":"The goal was to allow users to override system default using environment variables. So a separate `[systemdefaults]` (proposed by @durin42) becomes unnecessary.","author":"quark","id":"35160","dateCreated":"1518290016","dateModified":"1518290016"},{"type":"comment","comment":">>! In D2079#35154, @yuja wrote:\n>> I'm not sure if the precedence of settings is proper here.\n> \n> Perhaps it should be placed at the same level as `$EDITOR` (i.e. envrcitems)?\n\nI haven't bothered checking what that means for precedence, but I have an argument for giving NO_COLOR higher precedence than at least the system config: we set the system config for all our users at Google and some may want to override with NO_COLOR. I know someone (Jun? Yuya?) did some work related to precedence of EDITOR a while ago, but I don't remember what the goal of that was.\n","author":"martinvonz","id":"35157","dateCreated":"1518279995","dateModified":"1518279995"},{"type":"comment","comment":"> I'm not sure if the precedence of settings is proper here.\n\nPerhaps it should be placed at the same level as `$EDITOR` (i.e. envrcitems)?\n\nAnd can you update help\/color and environment?","author":"yuja","id":"35155","dateCreated":"1518265384","dateModified":"1518265384"},{"type":"comment","comment":"You might want to let run-tests.py drop NO_COLOR for tests.","author":"quark","id":"34711","dateCreated":"1518036117","dateModified":"1518036117"},{"type":"comment","comment":":+1: Hurrah for standards!","author":"lothiraldan","id":"34709","dateCreated":"1518024955","dateModified":"1518024955"},{"type":"accept","author":"lothiraldan","id":"34708","dateCreated":"1518024955","dateModified":"1518024955"},{"type":"update","diffId":"5298","author":"indygreg","id":"34702","dateCreated":"1518023858","dateModified":"1518023858"}],"dateCreated":"1518023858","dateModified":"1518697359","status":"Needs Revision"},{"id":"2259","callsign":"HG","title":"py3: use pycompat.bytestr to convert int\/bytes to bytes","author":"pulkit","summary":"The item value can be integer or can be bytes, so cannot use \"%d\" here.","testPlan":"","lineCount":"2","dependsOn":[],"reviewers":["yuja"],"ccs":["yuja","mercurial-devel"],"actions":[{"type":"comment","comment":"> The item value can be integer or can be bytes, so cannot use \"%d\" here.\n\nSounds like a bug. IIUC, the item should be a byte string because it is\npassed to `ui.progress(item=)`. Can you take a look?","author":"yuja","id":"37455","dateCreated":"1518614063","dateModified":"1518614063"},{"type":"reject","author":"yuja","id":"37454","dateCreated":"1518614063","dateModified":"1518614063"},{"type":"update","diffId":"5699","author":"pulkit","id":"37333","dateCreated":"1518607500","dateModified":"1518607500"}],"dateCreated":"1518607500","dateModified":"1518614064","status":"Needs Revision"},{"id":"2158","callsign":"HG","title":"py3: use raw string for open() mode","author":"indygreg","summary":" ","testPlan":"","lineCount":"8","dependsOn":[],"reviewers":[],"ccs":["yuja","pulkit","mercurial-devel"],"actions":[{"type":"comment","comment":"I wanna leave it with no prefix as possible. I think we'll have to bulk-replace\n`''`s to `b''`s at some point so we can leverage static analysis tools, where\n`r''` will be unneeded.","author":"yuja","id":"35990","dateCreated":"1518432079","dateModified":"1518432079"},{"type":"comment","comment":"Looks like a better approach will be to revert this change: https:\/\/www.mercurial-scm.org\/repo\/hg\/rev\/7c54917b31f6. I did that in favor of pycompat.open() and now I think it's better to revert that change. ","author":"pulkit","id":"35944","dateCreated":"1518420826","dateModified":"1518420826"},{"type":"update","diffId":"5459","author":"indygreg","id":"35745","dateCreated":"1518402936","dateModified":"1518402936"}],"dateCreated":"1518402936","dateModified":"1518432079","status":"Needs Review"},{"id":"2010","callsign":"HG","title":"check-commit: allow foo_bar naming in functions","author":"indygreg","summary":"nameswithallthewordssmashedtogetherarehardtoread.\nespeciallyifenglishisnotyourprimarylanguage.\n\nLet's align with the rest of the programming universe and\nallow_the_use_of_underscores_in_names.","testPlan":"","lineCount":"10","dependsOn":[],"reviewers":["pulkit","durin42"],"ccs":["av6","yuja","durin42","pulkit","mercurial-devel"],"actions":[{"type":"comment","comment":"As opposed to 11 words smashed together, function names in Mercurial usually are 2 or 3 words (if not just one), which is not at all bad. I think breaking consistency with the existing code base would be worse than making people get used to no-spaces style (the one that python's stdlib usually follows, in fact). So, FWIW, -1 from me.","author":"av6","id":"33903","dateCreated":"1517723192","dateModified":"1517723215"},{"type":"comment","comment":"I'm kinda -1 on this because of inconsistency with the 10-year codebase.","author":"yuja","id":"33887","dateCreated":"1517646932","dateModified":"1517646932"},{"type":"comment","comment":"I'm also +1 on this.","author":"durin42","id":"33697","dateCreated":"1517595429","dateModified":"1517595429"},{"type":"accept","author":"durin42","id":"33696","dateCreated":"1517595429","dateModified":"1517595429"},{"type":"comment","comment":"I am +1 on this.","author":"pulkit","id":"33694","dateCreated":"1517595331","dateModified":"1517595331"},{"type":"accept","author":"pulkit","id":"33693","dateCreated":"1517595331","dateModified":"1517595331"},{"type":"update","diffId":"5162","author":"indygreg","id":"33687","dateCreated":"1517595246","dateModified":"1517595246"}],"dateCreated":"1517595246","dateModified":"1517723192","status":"Needs Review"},{"id":"1937","callsign":"FBHGX","title":"p4fastimport: import purged files as empty files","author":"hollisb","summary":"P4 may preserve only the most recent copy of a file's data, automatically\n\"purging\" the file data from older changelists when a new one is submitted.\nThat data is gone, and can't be accessed even from P4.\n\nWhen importing such a file, we create an empty file so we can record that it\nexisted (e.g. which changelist introduced it).\n\nTherefore, we include it in the list of filelogs we import, and we record that\nchangelists modified the file too. The file just has empty contents in the\nolder changesets.","testPlan":"","lineCount":"56","dependsOn":[],"reviewers":["dsp"],"ccs":[],"actions":[{"type":"update","diffId":"5001","author":"hollisb","id":"32644","dateCreated":"1516757506","dateModified":"1516757506"}],"dateCreated":"1516757506","dateModified":"1516757576","status":"Needs Review"},{"id":"1936","callsign":"FBHGX","title":"p4fastimport: report more information about unsupported p4 file actions","author":"hollisb","summary":"No matter what, users and developers need this information to understand the\nproblem.","testPlan":"","lineCount":"3","dependsOn":[],"reviewers":["dsp"],"ccs":[],"actions":[{"type":"comment","comment":"I will have a followup patch to handle 'purge' file actions at some point, but it's not yet clear to me how to do it in the most useful and nonsurprising way. :-) In the meantime, this is a useful step.","author":"hollisb","id":"32642","dateCreated":"1516666023","dateModified":"1516666023"},{"type":"update","diffId":"5000","author":"hollisb","id":"32637","dateCreated":"1516665895","dateModified":"1516665895"}],"dateCreated":"1516665895","dateModified":"1516666023","status":"Needs Review"},{"id":"1649","callsign":"FBHGX","title":"infinitepush: add logic to support old clients on the basis of pushkey part","author":"pulkit","summary":"This patch adds logic to check whether a push is an infinitepush or not from\nan old client or client who don't have infinitepush client logic.","testPlan":"","lineCount":"60","dependsOn":[],"reviewers":["durham"],"ccs":[],"actions":[{"type":"comment","comment":"Status update: While adding tests, I found that I am reading the part twice, once for adding in bundlepart and once for the copied part. Since I can't read that twice, I need to do better here with the logic involved.","author":"pulkit","id":"31242","dateCreated":"1515664035","dateModified":"1515664064"},{"type":"inline","comment":"Can we add a comment saying this could require a lot of memory, and that we might want to move this to being stored on disk in the future?","replyTo":null,"isNewFile":"1","line":"1080","lineLength":"0","path":"infinitepush\/__init__.py","diffId":"4342","author":"durham","id":"30893","dateCreated":"1515514134","dateModified":"1515514134"},{"type":"inline","comment":"I'd make it clear that this will have performance impact.","replyTo":null,"isNewFile":"1","line":"1055","lineLength":"0","path":"infinitepush\/__init__.py","diffId":"4342","author":"durham","id":"30892","dateCreated":"1515514134","dateModified":"1515514134"},{"type":"inline","comment":"I'd rename this `supportoldclients`, since it's a server configuration.","replyTo":null,"isNewFile":"1","line":"1108","lineLength":"0","path":"infinitepush\/__init__.py","diffId":"4342","author":"durham","id":"30891","dateCreated":"1515514134","dateModified":"1515514134"},{"type":"comment","comment":"Could this have a test? Is there a config knob to disable bundle2 or force bundle1?","author":"durham","id":"30890","dateCreated":"1515514134","dateModified":"1515514134"},{"type":"reject","author":"durham","id":"30889","dateCreated":"1515514134","dateModified":"1515514134"},{"type":"update","diffId":"4342","author":"pulkit","id":"28287","dateCreated":"1512952476","dateModified":"1512952476"}],"dateCreated":"1512952476","dateModified":"1515664035","status":"Needs Revision"},{"id":"1779","callsign":"HG","title":"unionrepo: use named attributes on revlog index entries","author":"indygreg","summary":"This appears to be the last user of the offsets in core (as far as\nthe test harness was able to reveal).","testPlan":"","lineCount":"11","dependsOn":["1778"],"reviewers":["durin42"],"ccs":["durin42","mercurial-devel"],"actions":[{"type":"comment","comment":"I think we should hold off landing until the native C type is ready. I don't like shipping a performance regression on such low-level code without any other user benefit to offset it. The assumption here is that I'm the only one who will absorb pain from this series lingering. I'm willing to take that hit.","author":"indygreg","id":"31193","dateCreated":"1515640397","dateModified":"1515640397"},{"type":"comment","comment":"Happy to see these land, but waiting for any objections from others.","author":"durin42","id":"31088","dateCreated":"1515624896","dateModified":"1515624896"},{"type":"accept","author":"durin42","id":"31087","dateCreated":"1515624896","dateModified":"1515624896"},{"type":"update","diffId":"4640","author":"indygreg","id":"30186","dateCreated":"1514334979","dateModified":"1514334979"}],"dateCreated":"1514334979","dateModified":"1515640397","status":"Accepted"},{"id":"1778","callsign":"HG","title":"debugcommands: use named attributes on revlog index entries","author":"indygreg","summary":" ","testPlan":"","lineCount":"16","dependsOn":["1777"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"31084","dateCreated":"1515624896","dateModified":"1515624896"},{"type":"update","diffId":"4741","author":"indygreg","id":"30870","dateCreated":"1515227185","dateModified":"1515227185"},{"type":"update","diffId":"4639","author":"indygreg","id":"30177","dateCreated":"1514334977","dateModified":"1514334977"}],"dateCreated":"1514334977","dateModified":"1515624896","status":"Accepted"},{"id":"1777","callsign":"HG","title":"upgrade: use named attributes on revlog index entries","author":"indygreg","summary":" ","testPlan":"","lineCount":"6","dependsOn":["1776"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"31082","dateCreated":"1515624895","dateModified":"1515624895"},{"type":"update","diffId":"4638","author":"indygreg","id":"30168","dateCreated":"1514334975","dateModified":"1514334975"}],"dateCreated":"1514334974","dateModified":"1515624895","status":"Accepted"},{"id":"1776","callsign":"HG","title":"bundlerepo: use named attributes on revlog index entries","author":"indygreg","summary":" ","testPlan":"","lineCount":"4","dependsOn":["1775"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"31080","dateCreated":"1515624894","dateModified":"1515624894"},{"type":"update","diffId":"4637","author":"indygreg","id":"30159","dateCreated":"1514334972","dateModified":"1514334972"}],"dateCreated":"1514334972","dateModified":"1515624894","status":"Accepted"},{"id":"1775","callsign":"HG","title":"dagutil: use named attributes on revlog index entries","author":"indygreg","summary":"Continuing the transition away from tuples.","testPlan":"","lineCount":"20","dependsOn":["1774"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"31078","dateCreated":"1515624893","dateModified":"1515624893"},{"type":"update","diffId":"4636","author":"indygreg","id":"30149","dateCreated":"1514334970","dateModified":"1514334970"}],"dateCreated":"1514334970","dateModified":"1515624893","status":"Accepted"},{"id":"1774","callsign":"HG","title":"repoview: use named attributes on revlog index entries","author":"indygreg","summary":"We want to remove __getitem__ support from index entries. Port\na usage in repoview.","testPlan":"","lineCount":"2","dependsOn":["1773"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"31076","dateCreated":"1515624893","dateModified":"1515624893"},{"type":"update","diffId":"4740","author":"indygreg","id":"30869","dateCreated":"1515227182","dateModified":"1515227182"},{"type":"update","diffId":"4635","author":"indygreg","id":"30139","dateCreated":"1514334968","dateModified":"1514334968"}],"dateCreated":"1514334968","dateModified":"1515624893","status":"Accepted"},{"id":"1773","callsign":"HG","title":"revlog: use named attributes on revlog index entries","author":"indygreg","summary":"Now that we're using a dedicated type for index entries everywhere,\nwe can access fields in them by name instead of integer offset.\n\nThis change has a significant impact on performance on the Firefox\nrepository. (Values are from before refactor, parent commit, this\ncommit.)\n\n$ HGMODULEPOLICY=c hg perfrevset '::tip'\n! wall 0.755869 comb 0.750000 user 0.750000 sys 0.000000 (best of 13)\n! wall 1.011107 comb 1.010000 user 1.010000 sys 0.000000 (best of 10)\n! wall 0.864906 comb 0.860000 user 0.860000 sys 0.000000 (best of 11)\n\n$ HGMODULEPOLICY=py hg perfrevset '::tip'\n! wall 1.366281 comb 1.360000 user 1.360000 sys 0.000000 (best of 7)\n! wall 1.860068 comb 1.860000 user 1.860000 sys 0.000000 (best of 5)\n! wall 1.672432 comb 1.680000 user 1.670000 sys 0.010000 (best of 6)\n\nAnd with PyPy:\n\n! wall 0.163141 comb 0.160000 user 0.160000 sys 0.000000 (best of 56)\n! wall 0.154325 comb 0.160000 user 0.160000 sys 0.000000 (best of 54)\n! wall 0.155534 comb 0.160000 user 0.160000 sys 0.000000 (best of 53)\n\nThe C extension is ~15% slower than before the refactor. This is a\nbit unfortunate. That is a bit steep price to pay for more readable\ncode and the establishment of a formal interface for index entries.\n\nHowever, all is not lost! Because we're now using a custom type and\nconforming to a yet-to-be-formally-defined interface, there's nothing\npreventing us from implementing a backed-by-C custom type for revlog\nentries. This type could lazily resolve PyObject instances, which\nshould result in a massive performance boost for operations that\ndon't need to access all fields of the index entry. This is easier\ndone *after* we drop all uses of tuples and their API for referring\nto index entries.","testPlan":"","lineCount":"56","dependsOn":["1772"],"reviewers":["durin42"],"ccs":["durin42","yuja","mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"31074","dateCreated":"1515624892","dateModified":"1515624892"},{"type":"comment","comment":">>! In D1773#30628, @yuja wrote:\n>> However, it is quite large and I suspect it will take a bit more effort to finish it.\n> \n> Isn't that something like `dirstateTupleType`? If we had a native type, we wouldn't\n> need D1767, D1768, and probably D1769. If it can get rid of refcounting business,\n> reviewing this series will be much fun.\n\nI don't feel strongly about the approach. My bias is to land this whole series as-is. The \"lazy struct\" should be fairly easily done as a followup.\n\nThe refcounting of the type looks okay to me.","author":"durin42","id":"31061","dateCreated":"1515624811","dateModified":"1515624811"},{"type":"comment","comment":"> However, it is quite large and I suspect it will take a bit more effort to finish it.\n\nIsn't that something like `dirstateTupleType`? If we had a native type, we wouldn't\nneed D1767, D1768, and probably D1769. If it can get rid of refcounting business,\nreviewing this series will be much fun.","author":"yuja","id":"30628","dateCreated":"1515131007","dateModified":"1515131007"},{"type":"comment","comment":">>! In D1773#30283, @yuja wrote:\n> Not reviewed the series yet, but can't we just start with a custom type\n> backed by C so we don't have to import back IndexV1Entry from Python?\n\nThat's possible. I do have a work-in-progress patch to the C code to implement such a type.\n\nHowever, it is quite large and I suspect it will take a bit more effort to finish it. The end result of this series gets us into a state where the C type can be swapped in with minimal effort. I have a soft preference for implementing the type in C as a follow-up. But I can refactor the series if wanted. Or someone can review this series without accepting it and I'll add the C type later.","author":"indygreg","id":"30285","dateCreated":"1514487399","dateModified":"1514487399"},{"type":"comment","comment":"> Because we're now using a custom type and conforming to\n> a yet-to-be-formally-defined interface, there's nothing\n> preventing us from implementing a backed-by-C custom type\n> for revlog entries. \n\nNot reviewed the series yet, but can't we just start with a custom type\nbacked by C so we don't have to import back IndexV1Entry from Python?","author":"yuja","id":"30284","dateCreated":"1514469876","dateModified":"1514469876"},{"type":"update","diffId":"4634","author":"indygreg","id":"30129","dateCreated":"1514334966","dateModified":"1514334966"}],"dateCreated":"1514334966","dateModified":"1515624892","status":"Accepted"},{"id":"1772","callsign":"HG","title":"tests: port revlog index code to modern API","author":"indygreg","summary":"All revlog index parsers are now using our new type to\nrepresent index entries. We can now undo the custom type\ncomparison code we implemented to see us through the transition.\n\nIn addition, the code has been ported to use named attributes\ninstead of field offsets.","testPlan":"","lineCount":"48","dependsOn":["1771"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"31072","dateCreated":"1515624891","dateModified":"1515624891"},{"type":"update","diffId":"4739","author":"indygreg","id":"30867","dateCreated":"1515227180","dateModified":"1515227180"},{"type":"update","diffId":"4633","author":"indygreg","id":"30119","dateCreated":"1514334964","dateModified":"1514334964"}],"dateCreated":"1514334964","dateModified":"1515624891","status":"Accepted"},{"id":"1771","callsign":"HG","title":"revlog: pass index entry type into packentry() (API)","author":"indygreg","summary":"packentry() currently accepts tuples. Let's pass in index entry types\ninstead.\n\nThe censor extension was calling into this low-level code. It has\nbeen refactored to use our new type and to look up fields in index\nentries by name instead of offset.","testPlan":"","lineCount":"34","dependsOn":["1770"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"31070","dateCreated":"1515624891","dateModified":"1515624891"},{"type":"update","diffId":"4632","author":"indygreg","id":"30109","dateCreated":"1514334962","dateModified":"1514334962"}],"dateCreated":"1514334962","dateModified":"1515624891","status":"Accepted"},{"id":"1770","callsign":"HG","title":"revlog: use revlog index entry type for version 0 revlogs","author":"indygreg","summary":"The pure Python parser for version 1 revlogs uses it. Let's\nplug it into version 0 revlogs as well.","testPlan":"","lineCount":"9","dependsOn":["1769"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"31068","dateCreated":"1515624890","dateModified":"1515624890"},{"type":"update","diffId":"4631","author":"indygreg","id":"30099","dateCreated":"1514334959","dateModified":"1514334959"}],"dateCreated":"1514334959","dateModified":"1515624890","status":"Accepted"},{"id":"1769","callsign":"HG","title":"cext: use dedicated type for index entries","author":"indygreg","summary":"Now that we have a handle on our type to represent revlog index\nentries, let's use it.\n\nThis commit essentially consists of porting code from PyTuple to\nPyObject.\n\nAs part of porting the code, we now retrieve elements in the entry\ntype by named attribute instead of integer index.\n\nBefore, PyTuple_* APIs allowed us to retrieve a borrowed reference\nto PyObject elements. We can no longer do this via the PyObject_*\nAPIs for a type not implemented in C. This required extra refcount\nmanipulation in various places.\n\nThe C extension is now emitting the new type. And because various\ncode is still accessing index entry elements via __getitem__, we\nsee a performance regression in the Firefox repository:\n\n$ hg perfrevset '::tip'\n! wall 0.755869 comb 0.750000 user 0.750000 sys 0.000000 (best of 13)\n! wall 1.011107 comb 1.010000 user 1.010000 sys 0.000000 (best of 10)\n\nWe will claw back this regression in subsequent commits by accessing\nfields by name instead of index.\n\nThe version of the \"parsers\" C extension has been incremented to\nreflect the change in behavior.","testPlan":"","lineCount":"123","dependsOn":["1768"],"reviewers":["yuja","durin42"],"ccs":["yuja","mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"31067","dateCreated":"1515624889","dateModified":"1515624889"},{"type":"update","diffId":"4738","author":"indygreg","id":"30864","dateCreated":"1515227178","dateModified":"1515227178"},{"type":"inline","comment":"No error reported to the caller.\n\nAttributeError would be set automatically.","replyTo":null,"isNewFile":"1","line":"1788","lineLength":"0","path":"mercurial\/cext\/revlog.c","diffId":"4630","author":"yuja","id":"30618","dateCreated":"1515129734","dateModified":"1515129734"},{"type":"inline","comment":"I slightly prefer `goto bail` instead of `result = NULL; goto cleanup` pairs.","replyTo":null,"isNewFile":"1","line":"327","lineLength":"0","path":"mercurial\/cext\/revlog.c","diffId":"4630","author":"yuja","id":"30617","dateCreated":"1515129734","dateModified":"1515129734"},{"type":"inline","comment":"This is tricky, but looks okay. Alternatively, we could inc\/decref borrowed `obj`.","replyTo":null,"isNewFile":"1","line":"321","lineLength":"0","path":"mercurial\/cext\/revlog.c","diffId":"4630","author":"yuja","id":"30616","dateCreated":"1515129734","dateModified":"1515129734"},{"type":"inline","comment":"Perhaps AttributeError would be set by `PyObject_GetAttrString()`.","replyTo":null,"isNewFile":"1","line":"141","lineLength":"0","path":"mercurial\/cext\/revlog.c","diffId":"4630","author":"yuja","id":"30615","dateCreated":"1515129734","dateModified":"1515129734"},{"type":"reject","author":"yuja","id":"30614","dateCreated":"1515129734","dateModified":"1515129734"},{"type":"update","diffId":"4630","author":"indygreg","id":"30089","dateCreated":"1514334957","dateModified":"1514334957"}],"dateCreated":"1514334957","dateModified":"1515624889","status":"Needs Review"},{"id":"1768","callsign":"HG","title":"cext: obtain reference to index entry type","author":"indygreg","summary":"We recently introduced a dedicated type for index entries in the\npure Python implementation.\n\nThis commit tells the revlog C code about that type.\n\nWe register the type on the parsers module so it is exposed to\nimporters of the module.\n\nIn addition, we store a reference to the type on revlog index\ninstances. In theory, we could grab the reference from the\nmodule. However, this requires a fair bit of C code. It is easier\nto just re-import the original module and get a ref from there. This\nalso avoids problems with caching a type on a reload()d module.","testPlan":"","lineCount":"35","dependsOn":["1767"],"reviewers":["yuja","durin42"],"ccs":["yuja","mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"31066","dateCreated":"1515624888","dateModified":"1515624888"},{"type":"update","diffId":"4737","author":"indygreg","id":"30861","dateCreated":"1515227177","dateModified":"1515227177"},{"type":"inline","comment":"In theory, there could be a mismatch.\n\nIf this module is loaded, its `IndexV1Entry` will refer to whatever `mercurial.pure.parsers.IndexV1Entry` is. If `mercurial.pure.parsers` is then reloaded and its `IndexV1Entry` changes and a revlog index type is created, its `IndexV1Entry` will be the new one from the reloaded `mercurial.pure.parsers`.\n\nSince I plan to replace the Python-implemented type with a C backed type, I'm inclined to not care about this.","replyTo":"30611","isNewFile":"1","line":"2117","lineLength":"0","path":"mercurial\/cext\/revlog.c","diffId":"4629","author":"indygreg","id":"30845","dateCreated":"1515222660","dateModified":"1515222660"},{"type":"inline","comment":"Can we be sure that this `IndexV1Entry` is identical to `self->entrytype`\nafter reloading Python modules?","replyTo":null,"isNewFile":"1","line":"2117","lineLength":"0","path":"mercurial\/cext\/revlog.c","diffId":"4629","author":"yuja","id":"30611","dateCreated":"1515128067","dateModified":"1515128067"},{"type":"comment","comment":"> We hold off incrementing the version of the \"parsers\" extension\n> because nothing in core relies on the new API yet.\n\nIt's really minor nit, but we have to increment the version since new\ncompiled module doesn't work with old `pure.parsers` module, which\nprovides no IndexV1Entry type.","author":"yuja","id":"30610","dateCreated":"1515128067","dateModified":"1515128067"},{"type":"reject","author":"yuja","id":"30609","dateCreated":"1515128067","dateModified":"1515128067"},{"type":"update","diffId":"4629","author":"indygreg","id":"30079","dateCreated":"1514334955","dateModified":"1514334955"}],"dateCreated":"1514334955","dateModified":"1515624888","status":"Needs Review"},{"id":"1767","callsign":"HG","title":"cext: make nullentry a member of index types","author":"indygreg","summary":"An upcoming commit will change the type of index entries from tuple\nto something else. This type may not be initialized\/available at\nmodule init time.\n\nWe prepare for this future change by making null entries a per-instance\nvariable on the index type.\n\nIn theory, this does add some overhead (a new 8-tuple per index\ninstance). However, the overhead should be negligible compared to\nthe run-time overhead of creating an index entry for each revision\nin the index.","testPlan":"","lineCount":"19","dependsOn":["1766"],"reviewers":["yuja","durin42"],"ccs":["yuja","mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"31065","dateCreated":"1515624888","dateModified":"1515624888"},{"type":"update","diffId":"4736","author":"indygreg","id":"30859","dateCreated":"1515227176","dateModified":"1515227176"},{"type":"inline","comment":"This comment shouldn't be moved. We need to NULLify all attributes, but don't\nhave to create PyObject.","replyTo":null,"isNewFile":"1","line":"1876","lineLength":"0","path":"mercurial\/cext\/revlog.c","diffId":"4628","author":"yuja","id":"30606","dateCreated":"1515124415","dateModified":"1515124415"},{"type":"reject","author":"yuja","id":"30605","dateCreated":"1515124415","dateModified":"1515124415"},{"type":"update","diffId":"4628","author":"indygreg","id":"30069","dateCreated":"1514334953","dateModified":"1514334953"}],"dateCreated":"1514334953","dateModified":"1515624888","status":"Needs Review"},{"id":"1766","callsign":"HG","title":"cext: rename variable describing format for index entries","author":"indygreg","summary":"\"tuple_format\" is ambiguous.","testPlan":"","lineCount":"6","dependsOn":["1765"],"reviewers":["durin42"],"ccs":["mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"31063","dateCreated":"1515624887","dateModified":"1515624887"},{"type":"update","diffId":"4627","author":"indygreg","id":"30059","dateCreated":"1514334951","dateModified":"1514334951"}],"dateCreated":"1514334951","dateModified":"1515624887","status":"Accepted"},{"id":"1765","callsign":"HG","title":"parsers: use an attr-derived class for revlog index entries","author":"indygreg","summary":"Currently, revlog index entries are tuples.\n\nThis commit introduces a new type to represent revlog v1 index\nentries. The pure Python parser has been converted to use it.\n\nTo provide backwards compatibility with existing code (including the\nC extension), the type has a __getitem__ to facilitate index\nlookups. The intent is to remove this __getitem__ once all consumers\nare using named attributes.\n\nSince the pure Python index parser is now using our new type,\ntest-parseindex2.py had to be updated to deal with the type mismatch.\nOnce all index parsers are converted and the new type\/interface is\nubiquitous, we can restore the simplicity of test-parseindex2.py. We\nalso had to (temporarily) add \"# no-check-code\" to test-parseindex2.py\nto allow the import of the pure module. This will be removed once the C\nextension is ported.\n\nBecause consumers are going through an extra Python function call to\n__getitem__ to resolve attributes by index, this change regresses\nperformance on a simple DAG walking revset for the Firefox repository:\n\n$ HGMODULEPOLICY=py hg perfrevset '::tip'\n! wall 1.366281 comb 1.360000 user 1.360000 sys 0.000000 (best of 7)\n! wall 1.860068 comb 1.860000 user 1.860000 sys 0.000000 (best of 5)\n\nUsing PyPy, a major regression is not apparent:\n\n! wall 0.163141 comb 0.160000 user 0.160000 sys 0.000000 (best of 56)\n! wall 0.154325 comb 0.160000 user 0.160000 sys 0.000000 (best of 54)","testPlan":"","lineCount":"70","dependsOn":[],"reviewers":["yuja","durin42"],"ccs":["yuja","mercurial-devel"],"actions":[{"type":"accept","author":"durin42","id":"31062","dateCreated":"1515624886","dateModified":"1515624886"},{"type":"update","diffId":"4735","author":"indygreg","id":"30856","dateCreated":"1515227175","dateModified":"1515227175"},{"type":"inline","comment":"Perhaps you mean `list(map(...))` instead of a list of one element?","replyTo":null,"isNewFile":"1","line":"182","lineLength":"0","path":"tests\/test-parseindex2.py","diffId":"4626","author":"yuja","id":"30602","dateCreated":"1515124211","dateModified":"1515124211"},{"type":"reject","author":"yuja","id":"30601","dateCreated":"1515124211","dateModified":"1515124211"},{"type":"update","diffId":"4626","author":"indygreg","id":"30049","dateCreated":"1514334949","dateModified":"1514334949"}],"dateCreated":"1514334949","dateModified":"1515624886","status":"Needs Review"},{"id":"1714","callsign":"FBHGX","title":"distutils_rust: use atomic replace to install the .so file","author":"quark","summary":"Python loads .so using mmap. Rewriting a .so in-place when there is a Python\nprocess with that .so mmap-ed could crash that Python process. Therefore\nlet's avoid rewriting the destination .so in-place.\n\nThis was discovered by a txnclose Python hook I used to keep binaries\nup-to-date. It caused segfault after building the Rust extensions currently\nloaded by the Python process running the hook.","testPlan":"","lineCount":"5","dependsOn":[],"reviewers":["mbthomas"],"ccs":[],"actions":[{"type":"accept","author":"mbthomas","id":"30880","dateCreated":"1515440159","dateModified":"1515440159"},{"type":"update","diffId":"4529","author":"quark","id":"29324","dateCreated":"1513624671","dateModified":"1513624671"}],"dateCreated":"1513624671","dateModified":"1515440159","status":"Accepted"},{"id":"1828","callsign":"FBHGX","title":"morestatus: sync the extension with the logic moved to core","author":"pulkit","summary":"All the functionality of the morestatus extension was moved to core except the\nupdate part. The update part hacks around the update command and stores the\nargument passed before the update. This patch drops the code which was moved to\ncore, wraps the code in core to support the update state functionality. The\nsupport for update in extension is bit hacky as functions which returns helpful\nmessage for each state, does not take a repo argument in core.\n\nEven after this patch, the config options morestatus.show and\nmorestatus.skipstates works the same way as they used to be.\n\nThere are some newlines changes because of bit change in format of output in\ncore.","testPlan":"","lineCount":"209","dependsOn":[],"reviewers":[],"ccs":[],"actions":[{"type":"update","diffId":"4734","author":"pulkit","id":"30852","dateCreated":"1515225678","dateModified":"1515225678"}],"dateCreated":"1515225678","dateModified":"1515225678","status":"Needs Review"},{"id":"1821","callsign":"FBHGX","title":"treemanifest: subclass remotetreedatastore from generatingdatastore","author":"singhsrb","summary":"Introducing class generatingdatastore to represent all stores like\nremotetreedatastore which generate the trees when they are not found in the\nshared store, add them to the shared store, and thereafter, repeat the original\noperation on the shared store expecting it to succeed.","testPlan":"Ran all the tests.","lineCount":"50","dependsOn":["1820"],"reviewers":[],"ccs":[],"actions":[{"type":"update","diffId":"4721","author":"singhsrb","id":"30753","dateCreated":"1515197877","dateModified":"1515197877"}],"dateCreated":"1515197877","dateModified":"1515197910","status":"Needs Review"},{"id":"1820","callsign":"FBHGX","title":"treemanifest: refactor remotetreedatastore call hierarchy","author":"singhsrb","summary":"The plan is to eventually subclass remotetreedatastore and this\nrefactor sets us up nicely for that.","testPlan":"Ran all the tests.","lineCount":"13","dependsOn":[],"reviewers":[],"ccs":[],"actions":[{"type":"update","diffId":"4720","author":"singhsrb","id":"30747","dateCreated":"1515195853","dateModified":"1515195853"}],"dateCreated":"1515195853","dateModified":"1515197910","status":"Needs Review"},{"id":"1805","callsign":"FBHGX","title":"sparse: don't materialize sparse files during IMM","author":"phillco","summary":"There's no need to do this and not doing so speeds up IMM tremendously when the sparse profile had to be adjusted.","testPlan":"I ran an IMM before and after this change. Both succeeded but the later version was\nmuch faster.","lineCount":"10","dependsOn":[],"reviewers":["quark","durham"],"ccs":["durham"],"actions":[{"type":"update","diffId":"4701","author":"phillco","id":"30626","dateCreated":"1515130813","dateModified":"1515130813"},{"type":"inline","comment":"\"mutable\" is tricky since technically on the on-disk working context is mutable too.","replyTo":"30565","isNewFile":"1","line":"82","lineLength":"0","path":"hgext3rd\/fbsparse.py","diffId":"4691","author":"phillco","id":"30625","dateCreated":"1515130763","dateModified":"1515130763"},{"type":"inline","comment":"maybe \"working context\" would work","replyTo":"30564","isNewFile":"1","line":"82","lineLength":"0","path":"hgext3rd\/fbsparse.py","diffId":"4691","author":"phillco","id":"30565","dateCreated":"1515098574","dateModified":"1515098574"},{"type":"inline","comment":"Looks fine. It might be nice to come up with some better terminology. I think of \"working copy\" as always referring to the physical on disk one, so maybe we need some other term for a mutable context (of which working copy would be one type).","replyTo":null,"isNewFile":"1","line":"82","lineLength":"0","path":"hgext3rd\/fbsparse.py","diffId":"4691","author":"durham","id":"30564","dateCreated":"1515097649","dateModified":"1515097649"},{"type":"update","diffId":"4691","author":"phillco","id":"30563","dateCreated":"1515097415","dateModified":"1515097415"},{"type":"comment","comment":"I added a comment and also filed T24918295 for the wider task around allowing lazy filectx writes (but as Durham mentioned that's a strict improvement on the status quo).","author":"phillco","id":"30562","dateCreated":"1515097338","dateModified":"1515097338"},{"type":"update","diffId":"4690","author":"phillco","id":"30561","dateCreated":"1515097206","dateModified":"1515097206"},{"type":"comment","comment":"I'll add a comment.","author":"phillco","id":"30534","dateCreated":"1515093676","dateModified":"1515093676"},{"type":"comment","comment":"Discussed this in person. My concerns were that this would allow large action lists to propagate up and cause excess remotefilelog actions.  Phil's response was that A) this only applies to rebases for now, and any action in the rebase is going to require downloading anyway (when it the sparse profile gets temporarily expanded below), and B) any action on a in memory context should probably not be filtered by sparseness anyway.  I think the invariant in B is reasonable, and if we see it causing problems like A we can reevaluate how to fix that problem to meet invariant B.","author":"durham","id":"30533","dateCreated":"1515092819","dateModified":"1515092819"},{"type":"accept","author":"durham","id":"30532","dateCreated":"1515092819","dateModified":"1515092819"},{"type":"accept","author":"quark","id":"30529","dateCreated":"1515090793","dateModified":"1515090793"},{"type":"update","diffId":"4687","author":"phillco","id":"30524","dateCreated":"1515088883","dateModified":"1515088883"}],"dateCreated":"1515088883","dateModified":"1515130813","status":"Accepted"},{"id":"1743","callsign":null,"title":"p4fastimport: allow sync commits only when imported repository is up-to-date","author":"dsp","summary":"Previously we allowed a sync commit to happen at any time, but imported the\nfull repository to bring it into sync. This is slow for very large repositories\nand might be accidentally triggered. We change this behavior to only do a sync\ncommit if only the imported client changed but nothing else.","testPlan":"  cd tests\n  python ..\/..\/hg-crew\/tests\/run-tests.py test-p4fastimport-*.t test-check*.t","lineCount":"282","dependsOn":["1742"],"reviewers":["kaleyh"],"ccs":["kaleyh"],"actions":[{"type":"inline","comment":"that can be a good idea","replyTo":"30496","isNewFile":"1","line":"134","lineLength":"0","path":"tests\/test-p4fastimport-blobcommit.t","diffId":"4577","author":"kaleyh","id":"30509","dateCreated":"1515016570","dateModified":"1515016570"},{"type":"accept","author":"kaleyh","id":"30508","dateCreated":"1515016570","dateModified":"1515016570"},{"type":"inline","comment":"In case `p4fastimport` uses the new client-spec, it behaves like before: if a file changes it will happily included it. So what we want to do is: use the new-spec only after p4syncimport changed. Maybe in an upcoming change we can save information about the clientspec in Mercurial to automatically detect a change in the p4fastimport case","replyTo":"29849","isNewFile":"1","line":"134","lineLength":"0","path":"tests\/test-p4fastimport-blobcommit.t","diffId":"4577","author":"dsp","id":"30496","dateCreated":"1514922124","dateModified":"1514922124"},{"type":"request-review","author":"dsp","id":"30494","dateCreated":"1514921883","dateModified":"1514921883"},{"type":"inline","comment":"Yes, using the old clientspec it will be able to update it.","replyTo":"29849","isNewFile":"1","line":"134","lineLength":"0","path":"tests\/test-p4fastimport-blobcommit.t","diffId":"4577","author":"dsp","id":"30493","dateCreated":"1514921870","dateModified":"1514921870"},{"type":"inline","comment":"What client will it use if we fall back to `fast import` ? Since CL 5 touches files that are not in the current client spec, will `fast import` be able to update the repo to CL5?","replyTo":null,"isNewFile":"1","line":"134","lineLength":"0","path":"tests\/test-p4fastimport-blobcommit.t","diffId":"4577","author":"kaleyh","id":"29849","dateCreated":"1513982220","dateModified":"1513982220"},{"type":"reject","author":"kaleyh","id":"29848","dateCreated":"1513982220","dateModified":"1513982220"},{"type":"update","diffId":"4577","author":"dsp","id":"29750","dateCreated":"1513905522","dateModified":"1513905522"}],"dateCreated":"1513905522","dateModified":"1515016570","status":"Accepted"},{"id":"1787","callsign":null,"title":"rebase: turn off IMM if any rebased paths match a config","author":"phillco","summary":" ","testPlan":"","lineCount":"21","dependsOn":["1786"],"reviewers":["quark"],"ccs":["quark"],"actions":[{"type":"update","diffId":"4678","author":"phillco","id":"30456","dateCreated":"1514512002","dateModified":"1514512002"},{"type":"update","diffId":"4664","author":"phillco","id":"30334","dateCreated":"1514502802","dateModified":"1514502802"},{"type":"inline","comment":"Could you make this a regular expression?","replyTo":null,"isNewFile":"1","line":"914","lineLength":"0","path":"hgext\/rebase.py","diffId":"4663","author":"quark","id":"30332","dateCreated":"1514500915","dateModified":"1514500915"},{"type":"accept","author":"quark","id":"30331","dateCreated":"1514500915","dateModified":"1514500915"},{"type":"update","diffId":"4663","author":"phillco","id":"30317","dateCreated":"1514500517","dateModified":"1514500517"}],"dateCreated":"1514500517","dateModified":"1514512002","status":"Accepted"},{"id":"1786","callsign":null,"title":"rebase: extract _shoulddisableimm for clarity","author":"phillco","summary":" ","testPlan":"","lineCount":"41","dependsOn":["1785"],"reviewers":["quark"],"ccs":[],"actions":[{"type":"accept","author":"quark","id":"30328","dateCreated":"1514500793","dateModified":"1514500793"},{"type":"update","diffId":"4662","author":"phillco","id":"30310","dateCreated":"1514500514","dateModified":"1514500514"}],"dateCreated":"1514500514","dateModified":"1514500793","status":"Accepted"},{"id":"1785","callsign":null,"title":"rebase: log why we didn't use IMM","author":"phillco","summary":"In preparation for adding more conditions that will disable IMM, let's make\nthis framework more extensible and log the reason IMM wasn't used.\n\nThe `whynotimm` variable name was inspired by the `why_not_eden` name in\n`hg\/eden\/__init__.py`.","testPlan":"","lineCount":"26","dependsOn":[],"reviewers":["quark"],"ccs":[],"actions":[{"type":"inline","comment":"same here","replyTo":null,"isNewFile":"1","line":"996","lineLength":"0","path":"hgext\/rebase.py","diffId":"4661","author":"quark","id":"30326","dateCreated":"1514500758","dateModified":"1514500758"},{"type":"inline","comment":"+3 spaces","replyTo":null,"isNewFile":"1","line":"798","lineLength":"0","path":"hgext\/rebase.py","diffId":"4661","author":"quark","id":"30325","dateCreated":"1514500758","dateModified":"1514500758"},{"type":"accept","author":"quark","id":"30324","dateCreated":"1514500758","dateModified":"1514500758"},{"type":"update","diffId":"4661","author":"phillco","id":"30305","dateCreated":"1514498645","dateModified":"1514498645"}],"dateCreated":"1514498645","dateModified":"1514500758","status":"Accepted"},{"id":"1753","callsign":"HG","title":"streamclone: preserve remote phases (issue5648)","author":"indygreg","summary":"As the tests added by the previous changeset demonstrate, stream clones\nwere never taught about the existence of phases. As a result, all\nchangesets applied via stream clone were made public, even if they\nwere draft (or even secret) on the remote.\n\nThis commit integrates phases knowledge into stream clones.\n\nAs the inline comment explains, because there are scenarios where\nwe can't disambiguate phases support, the resulting phase data may\nnot match the server exactly. However, I believe the remaining\nareas of buggy behavior are confined to secret changesets. Secret\nchangesets aren't common. And transferring them via stream clones\neither requires a buggy Mercurial server or a specially-configured\nserver. I think it will require stream clone support in bundle2\nbefore we can properly handle secret phases for stream clones.\n\nIt \/might\/ be acceptable to force all local changesets to secret\nand then promote to draft or public from remote phases data. However,\nI was having trouble implementing this. Existing phases code doesn't\nseem geared towards supporting this scenario and I didn't want to\nincur a lot of extra work to refactor the phases code. Perfect is\nthe enemy of good: I'm inclined to tackle secret phase as part of\nsupporting stream clones in bundle2.\n\nThis change introduces a `roots(all())` revset, which will require\na linear scan of the changelog during clone bundles. On a Firefox\nrepo, this may slow down stream clones by ~1s. Again, support for\nstream clones in bundle2 should help here.\n\n.. fix::\n\n   Draft changeset phase is now preserved during a stream clone.\n\n   Before, all changesets pulled via a stream clone would have a\n   public phase, even if they were draft on the remote and the remote\n   was non-publishing.","testPlan":"","lineCount":"64","dependsOn":["1752"],"reviewers":["yuja"],"ccs":["yuja","mercurial-devel"],"actions":[{"type":"inline","comment":"Perhaps the remote sends back actual phases data even if it is a publishing server.\nThat's probably why `exchange._pullapplyphases()` checks 'publishing' flag.","replyTo":"30215","isNewFile":"1","line":"191","lineLength":"0","path":"mercurial\/streamclone.py","diffId":"4603","author":"yuja","id":"30280","dateCreated":"1514466877","dateModified":"1514466877"},{"type":"inline","comment":"Maybe.\n\nMy reasoning here was that if the remote advertises phases data, then it is phases aware and the code below will ensure phases are adjusted appropriately. We certainly could special case the mixed phase + publishing case and skip all phase adjustment.","replyTo":"29973","isNewFile":"1","line":"191","lineLength":"0","path":"mercurial\/streamclone.py","diffId":"4603","author":"indygreg","id":"30215","dateCreated":"1514399658","dateModified":"1514399658"},{"type":"inline","comment":"It seems odd that `remotephases` data aren't used at all, but\nI'm not sure.","replyTo":null,"isNewFile":"1","line":"200","lineLength":"0","path":"mercurial\/streamclone.py","diffId":"4603","author":"yuja","id":"29974","dateCreated":"1514293502","dateModified":"1514293502"},{"type":"inline","comment":"Shouldn't we test `remotephases.get('publishing', False)` for\nmixed-phase + publishing server?\n\n","replyTo":null,"isNewFile":"1","line":"191","lineLength":"0","path":"mercurial\/streamclone.py","diffId":"4603","author":"yuja","id":"29973","dateCreated":"1514293502","dateModified":"1514293502"},{"type":"reject","author":"yuja","id":"29972","dateCreated":"1514293502","dateModified":"1514293502"},{"type":"update","diffId":"4603","author":"indygreg","id":"29886","dateCreated":"1514138810","dateModified":"1514138810"}],"dateCreated":"1514138810","dateModified":"1514466877","status":"Needs Revision"},{"id":"1784","callsign":"FBHGX","title":"fbamend: autorestack children for simple rebases","author":"marco.leogrande","summary":"Today, if you run 'hg amend' while in the middle of a stack of\ncommits, it amends the commit and leaves the top of the stack\nbehind. Users then have to run 'hg amend --restack' or 'hg restack' to\nbring the old top of the stack onto the new version of the commit.\n\nAdd heuristic code so that an automatic restack can be performed if\namending a commit did not change its manifest, nor did it leave behind\na dirty working copy. This optimization can be avoided by providing\nthe --no-rebase commandline option.\n\nFix the existing tests that relied on the previous behaviour (so their\noriginal intent is kept), and add new ones to cover the new heuristic.","testPlan":"","lineCount":"107","dependsOn":[],"reviewers":["quark"],"ccs":["quark"],"actions":[{"type":"comment","comment":"Great!","author":"quark","id":"30274","dateCreated":"1514435622","dateModified":"1514435622"},{"type":"accept","author":"quark","id":"30273","dateCreated":"1514435622","dateModified":"1514435622"},{"type":"update","diffId":"4657","author":"marco.leogrande","id":"30271","dateCreated":"1514428200","dateModified":"1514428200"},{"type":"inline","comment":"`debugbuilddag` creates empty commits that requires explicit \"add\" if you changed \"b\". That's doable by using `hg amend -A b` here.","replyTo":null,"isNewFile":"1","line":"899","lineLength":"0","path":"tests\/test-fbamend-restack.t","diffId":"4656","author":"quark","id":"30267","dateCreated":"1514426580","dateModified":"1514426580"},{"type":"inline","comment":"`hg debugbuilddag '+3'` is a more efficient way to create multiple commits. You can use `hg update 'tip^'` in the next line to update to the second commit. That said, I'm fine with just using `mkcommit`.","replyTo":null,"isNewFile":"1","line":"873","lineLength":"2","path":"tests\/test-fbamend-restack.t","diffId":"4656","author":"quark","id":"30266","dateCreated":"1514426580","dateModified":"1514426580"},{"type":"inline","comment":"Since this is user-facing and not everyone knows what \"manifest\" is. It's more friendly to avoid the terminology. Power users may want to disable the feature (ex. if they want to amend multiple commits and then just use restack once to fix them all) but it's not obvious how. Maybe something like \"(auto-rebasing descendants, use --no-rebase to disable this)\"\n\nNote \"children\" and \"descendants\" are different in hg. See `hg help revset` for details.\n","replyTo":null,"isNewFile":"1","line":"278","lineLength":"1","path":"hgext3rd\/fbamend\/__init__.py","diffId":"4656","author":"quark","id":"30265","dateCreated":"1514426580","dateModified":"1514426580"},{"type":"comment","comment":"This looks pretty good. I just have a small request about the user-facing message.","author":"quark","id":"30264","dateCreated":"1514426580","dateModified":"1514426580"},{"type":"reject","author":"quark","id":"30263","dateCreated":"1514426580","dateModified":"1514426580"},{"type":"update","diffId":"4656","author":"marco.leogrande","id":"30258","dateCreated":"1514424219","dateModified":"1514424219"}],"dateCreated":"1514424219","dateModified":"1514435622","status":"Accepted"},{"id":"1752","callsign":"HG","title":"tests: show phases aren't preserved part of streaming clone (issue5648)","author":"indygreg","summary":"draft and secret changesets are promoted to public changesets when\ntransferred via a stream clone. That's not desirable behavior. Add\ntests demonstrating it.\n\nAs far as I can tell, this buggy behavior has been present since\nphases were introduced. Stream clone predated phases (1.7 versus 1.9)\nand stream clone was never taught to be phase aware.","testPlan":"","lineCount":"65","dependsOn":[],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"4602","author":"indygreg","id":"29880","dateCreated":"1514138808","dateModified":"1514138808"}],"dateCreated":"1514138808","dateModified":"1514138810","status":"Needs Review"},{"id":"1742","callsign":null,"title":"p4fastimport: return the last changelist number that affects the client view","author":"dsp","summary":"Previously we returned the last changelist number of any change a\nclient can see. This includes changes outside of the files specified in the\nview. We now return the last changelist number of files affected by the client\nview.","testPlan":"python ..\/..\/hg-crew\/tests\/run-tests.py test-check* test-p4fastimport*","lineCount":"3","dependsOn":["1741"],"reviewers":["quark"],"ccs":[],"actions":[{"type":"accept","author":"quark","id":"29766","dateCreated":"1513913174","dateModified":"1513913174"},{"type":"update","diffId":"4576","author":"dsp","id":"29741","dateCreated":"1513905502","dateModified":"1513905502"}],"dateCreated":"1513905502","dateModified":"1513913174","status":"Accepted"},{"id":"1741","callsign":null,"title":"p4fastimport: fix symlink test on MacOS","author":"dsp","summary":"On MacOS the resulting symlinks do have a differnt mode and a slightly\ndifferent formatting due to the use of BSD ls.","testPlan":"Ran\n\n  python ..\/..\/hg-crew\/tests\/run-tests.py test-check* test-p4fastimport*\n\nand see it fail on MacOS. After the fix ran it again and see it worked","lineCount":"4","dependsOn":[],"reviewers":["quark"],"ccs":[],"actions":[{"type":"accept","author":"quark","id":"29764","dateCreated":"1513913128","dateModified":"1513913128"},{"type":"update","diffId":"4575","author":"dsp","id":"29736","dateCreated":"1513905500","dateModified":"1513905500"}],"dateCreated":"1513905500","dateModified":"1513913128","status":"Accepted"},{"id":"1621","callsign":"HG","title":"transaction: encodes tuples in changes['phases'] as 4 bit ints","author":"joerg.sonnenberger","summary":"changes['phases'] contains tuples for the phase transitions, which\nprovides a non-trivial overhead in Python. Replace this with a Python\ndictionary containing 4 bit integers. Provide accessors to centralize\nchanges['phases'] procsessing in one place.\n\nOn average, this improves user time for issue5691 by 1% and peak RSS by\n3MB.","testPlan":"","lineCount":"78","dependsOn":[],"reviewers":["quark"],"ccs":["yuja","quark","mercurial-devel"],"actions":[{"type":"comment","comment":"When thinking about it again, I believe providing a way to disable `tr.chanages['phases']` is a better way to solve this issue. If we cannot disable that automatically from what hooks are used, we can still provide a config option to disable that.","author":"quark","id":"29293","dateCreated":"1513404110","dateModified":"1513404110"},{"type":"resign","author":"quark","id":"29292","dateCreated":"1513404110","dateModified":"1513404110"},{"type":"comment","comment":">>! In D1621#28387, @joerg.sonnenberger wrote:\n> The bitset version has shown already that optimizing this is worthwhile and can eliminate up to 10% of the total size of the transaction object.\n\nIIRC, Jun said a plain intbitset wouldn't be optimal for all scenarios, and I agree\nwith that. Also, it isn't a standard Python library.\n\nI think the current way of recording phase change is too ad-hoc if it does iterate\nall incoming nodes to build `(old, new)` phases pairs. Instead, can't we reconstruct\nthem from DAG + old\/new phase roots, for example?\n\nIf the current `tr.changes['phases']` is only used by hooks and third-party\nextensions, it can be opt-in feature.","author":"yuja","id":"28507","dateCreated":"1513084873","dateModified":"1513084873"},{"type":"comment","comment":"The bitset version has shown already that optimizing this is worthwhile and can eliminate up to 10% of the total size of the transaction object.\n\nThe change as is primarily gets the interfaces in place, so that outside code can be adjusted.","author":"joerg.sonnenberger","id":"28387","dateCreated":"1513015530","dateModified":"1513015530"},{"type":"comment","comment":">>! In D1621#28285, @joerg.sonnenberger wrote:\n> This version should be committable. It introduces the necessary API for isolating further changes and gives a small improvement already. I'll be looking at provider a memory and time efficient sparce vector next, that's where the real benefits will be.\n\nI think it's better to hold this change off until we can get a real performance or\nmemory-space win. The current number, \"peak RSS by 3MB\", won't buy much\ncompared to the API complexity this patch will introduce.\n\nRegarding the API, I think `tr.changes[key]` is the source of the problem. It allows\nfree read\/write access to arbitrary data, which is hard to gradually improve the\nunderlying data structure or algorithm.\n\nNoob quesiton. Can't we completely switch of `tr.changes['phases']` when it's\nunnecessary?","author":"yuja","id":"28324","dateCreated":"1513002345","dateModified":"1513002345"},{"type":"comment","comment":"This version should be committable. It introduces the necessary API for isolating further changes and gives a small improvement already. I'll be looking at provider a memory and time efficient sparce vector next, that's where the real benefits will be.","author":"joerg.sonnenberger","id":"28285","dateCreated":"1512940391","dateModified":"1512940391"},{"type":"update","diffId":"4341","author":"joerg.sonnenberger","id":"28282","dateCreated":"1512940262","dateModified":"1512940262"},{"type":"comment","comment":"> Original code: 485s \/ 584MB\n> Set version: 501s \/ 582MB\n> intbitset: 478s \/ 563MB\n\nI'm not sure how to read this. Do we only get -2\/584MB better space\nconsumption against the original simplest implementation?","author":"yuja","id":"27976","dateCreated":"1512825094","dateModified":"1512825094"},{"type":"comment","comment":"That said, I think the usage of list of sets is smart and this change looks good to me.","author":"quark","id":"27917","dateCreated":"1512779148","dateModified":"1512779148"},{"type":"accept","author":"quark","id":"27916","dateCreated":"1512779148","dateModified":"1512779148"},{"type":"inline","comment":"Since this hook is the only user of `tr.changes['phases']`. I'd suggest disabling calculating `tr.changes['phases']` if the hook does not exist.","replyTo":null,"isNewFile":"1","line":"1243","lineLength":"0","path":"mercurial\/localrepo.py","diffId":"4278","author":"quark","id":"27915","dateCreated":"1512778979","dateModified":"1512778979"},{"type":"comment","comment":"This is a Proof-of-Concept and with a follow-up in D1622 to use a C implementation for the bitset.\n\nOriginal code: 485s \/ 584MB\nSet version: 501s \/ 582MB\nintbitset: 478s \/ 563MB\n\n","author":"joerg.sonnenberger","id":"27912","dateCreated":"1512778236","dateModified":"1512778236"},{"type":"update","diffId":"4278","author":"joerg.sonnenberger","id":"27897","dateCreated":"1512777987","dateModified":"1512777987"}],"dateCreated":"1512777987","dateModified":"1513404110","status":"Needs Review"},{"id":"1518","callsign":"FBHGX","title":"nodemap: move some methods to changelog utils","author":"quark","summary":"Those methods and constants will be reused by another module.","testPlan":"","lineCount":"68","dependsOn":[],"reviewers":["jsgf"],"ccs":["jsgf"],"actions":[{"type":"update","diffId":"4461","author":"quark","id":"28995","dateCreated":"1513307716","dateModified":"1513307716"},{"type":"update","diffId":"4455","author":"quark","id":"28981","dateCreated":"1513304499","dateModified":"1513304499"},{"type":"inline","comment":"```\nif rev > i32::MAX as u64 {\n```\n\nYou may need to `use std::i32;`","replyTo":null,"isNewFile":"1","line":"13","lineLength":"0","path":"rust\/indexes\/src\/utils.rs","diffId":"3880","author":"jsgf","id":"27451","dateCreated":"1512667504","dateModified":"1512667504"},{"type":"inline","comment":"(driveby) You can use `b'a'` to get a character value as a `u8`. so `b'a'...b'f' => ch - b'a',`, etc","replyTo":null,"isNewFile":"1","line":"190","lineLength":"2","path":"rust\/indexes\/src\/nodemap.rs","diffId":"3880","author":"jsgf","id":"27450","dateCreated":"1512667504","dateModified":"1512667504"},{"type":"reject","author":"jsgf","id":"27449","dateCreated":"1512667504","dateModified":"1512667504"},{"type":"update","diffId":"3880","author":"quark","id":"25601","dateCreated":"1511807450","dateModified":"1511807450"}],"dateCreated":"1511807450","dateModified":"1513385306","status":"Needs Review"},{"id":"1521","callsign":"FBHGX","title":"clindex: use childmap to calculate phases","author":"quark","summary":"Calculating phases is O(len(changelog)) since it scans revisions starting\nwith min(draft()) (unfiltered). Use the childmap so it is actually\nO(len(not public())).\n\nThis saves 100+ms for a fbsource repo created in late 2015.\n\nBesides, the old phases code build a O(len(changelog)) Python list for\nrev -> phase queries. This patch does not build such list and saves\nO(len(changelog)) memory.","testPlan":"With verify mode on, run core hg tests and make sure there is no\ninconsistent cases with the original phases implementation.","lineCount":"38","dependsOn":["1520"],"reviewers":[],"ccs":[],"actions":[{"type":"update","diffId":"4458","author":"quark","id":"28985","dateCreated":"1513304505","dateModified":"1513304505"},{"type":"update","diffId":"3883","author":"quark","id":"25627","dateCreated":"1511807456","dateModified":"1511807456"}],"dateCreated":"1511807456","dateModified":"1513385283","status":"Needs Review"},{"id":"1703","callsign":"HG","title":"logtoprocess: connect all fds to \/dev\/null to avoid bad interaction with pager","author":"lothiraldan","summary":"We detected that pager is waiting for log-to-process script to finish, which\nis annoying when adding a script on commandfinish that does an HTTP push.\n\nThere seems to be no workaround on the script side and it will make the\nbehavior on Linux\/MacOS closer to the Windows behavior.\n\nThe drawback is that it makes the related tests more flaky as log-to-process\noutputs are now really asynchronous.\n\nIf it's considered a BC change, another option would be to add a config option\nfor this new behavior. I personally think that the different behavior between\nWindows and Linux is confusing and that it's a bug I would be fine with a new\nconfig option.","testPlan":"","lineCount":"15","dependsOn":["1702"],"reviewers":[],"ccs":["yuja","mercurial-devel"],"actions":[{"type":"inline","comment":"Better to close these file objects in the parent process, though that wouldn't matter.\n\nAs we only need a file descriptor, `os.open(os.devnull, os.O_RDWR)` can be used.","replyTo":null,"isNewFile":"1","line":"94","lineLength":"0","path":"hgext\/logtoprocess.py","diffId":"4471","author":"yuja","id":"29108","dateCreated":"1513347083","dateModified":"1513347083"},{"type":"update","diffId":"4471","author":"lothiraldan","id":"29050","dateCreated":"1513332929","dateModified":"1513332929"}],"dateCreated":"1513332929","dateModified":"1513347083","status":"Needs Review"},{"id":"1702","callsign":"HG","title":"logtoprocess: add a test to show pager and ltp bad interaction","author":"lothiraldan","summary":"The next changeset will fix the bad interaction between the pager and log-to-\nprocess. Add a test first to make the next changeset easier to understand.\n\nXXX-REVIEW: I tried making the test file compatible with chg but it seems\nflaky. The output is different when running only this test without parallelism\n(rt test-logtoprocess.t) vs running all tests with parallelism (rt -j 4).\nCould someone with more chg experience gives me a hand stabilizing this\ntest file?","testPlan":"","lineCount":"86","dependsOn":["1701"],"reviewers":["yuja"],"ccs":["yuja","mercurial-devel"],"actions":[{"type":"inline","comment":"IIUC, here `command.log` might not be written yet because the\nlogging process runs asynchronously.","replyTo":null,"isNewFile":"1","line":"50","lineLength":"0","path":"tests\/test-logtoprocess.t","diffId":"4470","author":"yuja","id":"29105","dateCreated":"1513346740","dateModified":"1513346740"},{"type":"comment","comment":"> XXX-REVIEW: I tried making the test file compatible with chg but it seems\n> flaky. The output is different when running only this test without parallelism\n> (rt test-logtoprocess.t) vs running all tests with parallelism (rt -j 4).\n\nCan you at least disable the broken tests on chg?","author":"yuja","id":"29104","dateCreated":"1513346740","dateModified":"1513346740"},{"type":"reject","author":"yuja","id":"29103","dateCreated":"1513346740","dateModified":"1513346740"},{"type":"update","diffId":"4470","author":"lothiraldan","id":"29040","dateCreated":"1513332925","dateModified":"1513332925"}],"dateCreated":"1513332925","dateModified":"1513346740","status":"Needs Revision"},{"id":"1701","callsign":"HG","title":"logtoprocess: add the possibility to not popup a console","author":"lothiraldan","summary":"Starting a subprocess on Windows with `DETACHED_PROCESS |\nCREATE_NEW_PROCESS_GROUP` as startup info means starting a console GUI window\nin almost all cases. If we add a logtoprocess for commandfinish and the prompt\ndo some hg calls, it will create a short-lived GUI window which is annoying\nfor end-users.\n\nCreating a subprocess with `CREATE_CONSOLE` instead seems to no start a\nconsole window but could have other side-effects so it kept under a config\nknob.","testPlan":"","lineCount":"13","dependsOn":["1700"],"reviewers":[],"ccs":["yuja","mercurial-devel"],"actions":[{"type":"inline","comment":"Can we add `CREATE_NEW_PROCESS_GROUP` ?","replyTo":null,"isNewFile":"1","line":"66","lineLength":"0","path":"hgext\/logtoprocess.py","diffId":"4468","author":"yuja","id":"29101","dateCreated":"1513346507","dateModified":"1513346507"},{"type":"comment","comment":"> but could have other side-effects so it kept under a config knob.\n\nMy two cents, `experimental.*` might be better if the config knob exists as\na workaround for an unknown bug.","author":"yuja","id":"29100","dateCreated":"1513346507","dateModified":"1513346507"},{"type":"update","diffId":"4468","author":"lothiraldan","id":"29029","dateCreated":"1513332660","dateModified":"1513332660"}],"dateCreated":"1513332660","dateModified":"1513346507","status":"Needs Review"},{"id":"1427","callsign":"HG","title":"logtoprocess: add a test to show pager and ltp bad interaction","author":"lothiraldan","summary":"The next changeset will fix the bad interaction between the pager and log-to-\nprocess. Add a test first to make the next changeset easier to understand.\n\nXXX-REVIEW: I tried making the test file compatible with chg but it seems\nflaky. The output is different when running only this test without parallelism\n(rt test-logtoprocess.t) vs running all tests with parallelism (rt -j 4).\nCould someone with more chg experience gives me a hand stabilizing this\ntest file?","testPlan":"","lineCount":"86","dependsOn":["1426"],"reviewers":["yuja"],"ccs":["yuja","durin42","quark","mercurial-devel"],"actions":[{"type":"inline","comment":"Can't we just append to the log file and sort the content?","replyTo":"29060","isNewFile":"1","line":"33","lineLength":"0","path":"tests\/test-logtoprocess.t","diffId":"3916","author":"yuja","id":"29062","dateCreated":"1513342276","dateModified":"1513342276"},{"type":"inline","comment":"I tried making the file path unique for each call (by adding `\\$(date +%s)`) but it doesn't seems to change anything in the stability unfortunately :(","replyTo":"26011","isNewFile":"1","line":"33","lineLength":"0","path":"tests\/test-logtoprocess.t","diffId":"3916","author":"lothiraldan","id":"29060","dateCreated":"1513335714","dateModified":"1513335714"},{"type":"inline","comment":"This file may be overwritten asynchronously by more than one\nlogging processes. That's probably why the test result is unstable\nespecially with chg.","replyTo":null,"isNewFile":"1","line":"33","lineLength":"0","path":"tests\/test-logtoprocess.t","diffId":"3916","author":"yuja","id":"26011","dateCreated":"1511960528","dateModified":"1511960528"},{"type":"reject","author":"yuja","id":"26010","dateCreated":"1511960528","dateModified":"1511960528"},{"type":"update","diffId":"3916","author":"lothiraldan","id":"25832","dateCreated":"1511901401","dateModified":"1511901401"},{"type":"comment","comment":"Could someone with chg experience give me a hand or a pointer why this test might be flaky with chg?","author":"lothiraldan","id":"25820","dateCreated":"1511900907","dateModified":"1511900907"},{"type":"comment","comment":"Jun, any idea why this might be chg-broken?","author":"durin42","id":"24127","dateCreated":"1510957537","dateModified":"1510957537"},{"type":"update","diffId":"3529","author":"lothiraldan","id":"23548","dateCreated":"1510760661","dateModified":"1510760661"}],"dateCreated":"1510760661","dateModified":"1513342276","status":"Needs Revision"},{"id":"1426","callsign":"HG","title":"logtoprocess: add the possibility to not start a shell","author":"lothiraldan","summary":"Starting a shell on Windows means starting a console GUI window in almost all\ncases. If we add a logtoprocess for commandfinish and the prompt do some hg\ncalls, it will create a short-lived GUI window which is annoying for end-\nusers.","testPlan":"","lineCount":"14","dependsOn":[],"reviewers":["mbthomas"],"ccs":["mharbison72","durin42","yuja","mbthomas","mercurial-devel"],"actions":[{"type":"comment","comment":"I've resent as https:\/\/phab.mercurial-scm.org\/D1700 because I couldn't get phabsend to reproduce my linear updated stack on phabricator :(","author":"lothiraldan","id":"29059","dateCreated":"1513333033","dateModified":"1513333033"},{"type":"comment","comment":"It works when passing `CREATE_NEW_CONSOLE` instead of `DETACHED_PROCESS | CREATE_NEW_PROCESS_GROUP`.\n\nMy concern now is about changing the behavior, do we want to keep the improvement under a config knob.","author":"lothiraldan","id":"28793","dateCreated":"1513271022","dateModified":"1513271022"},{"type":"comment","comment":"Sorry, I'm not sure what the context is here.  Is this all to resolve %FOO% type variables?  I don't know of any library to do this either.\n\nEarlier this year I tried fixing the logtoprocess test, which was leaving out chunks of output on Windows.  The only notes I left were that the output showed up if DETACHED_PROCESS was removed, but then it flickered a bunch of cmd.exe windows.  IDK how that works, so eventually I gave up and required no-windows.  That said, Yuya's code snippet seems to work for me on Win7 in a python shell.\n\nI don't have time to test tonight, but maybe this will help?  I vaguely remember having luck with SW_HIDE.\n\n  https:\/\/www.codeproject.com\/Articles\/2537\/Running-console-applications-silently\n\n","author":"mharbison72","id":"28773","dateCreated":"1513232248","dateModified":"1513232248"},{"type":"comment","comment":">>! In D1426#28695, @lothiraldan wrote:\n> Without `DETACHED_PROCESS`, no console is spawned either with `shell=False` or `shell=True` **but** Mercurial will wait for the log-to-process script to finish before ending.\n\nNo idea why, but a simple python script running `subprocess.Popen()` exits\nimmediately on my utterly old Windows XP.\n\n```\nimport subprocess\nimport sys\ncreationflags = subprocess.CREATE_NEW_PROCESS_GROUP\nsubprocess.Popen('%s -c \"import time; time.sleep(10)\"' % sys.executable,\n                 shell=True, creationflags=creationflags)\n```\n\n> So I would say to keep `DETACHED_PROCESS`.\n> \n> The real question is how to do variables expansion with `shell=False`, I don't think that Python has a stdlib function for doing that.\n\nEmulating cmd.exe behavior wouldn't be easy. If we had to set `shell=False`,\nwe'd probably better to document the Windows-specific issue.\n\n@mharbison72, do you have any idea?","author":"yuja","id":"28731","dateCreated":"1513173998","dateModified":"1513173998"},{"type":"comment","comment":"Without `DETACHED_PROCESS`, no console is spawned either with `shell=False` or `shell=True` **but** Mercurial will wait for the log-to-process script to finish before ending.\n\nSo I would say to keep `DETACHED_PROCESS`.\n\nThe real question is how to do variables expansion with `shell=False`, I don't think that Python has a stdlib function for doing that.","author":"lothiraldan","id":"28695","dateCreated":"1513155805","dateModified":"1513155805"},{"type":"comment","comment":"I've access to a Windows machine this week, so you can expect a comment or new version before the end of the week.","author":"lothiraldan","id":"28506","dateCreated":"1513068206","dateModified":"1513068206"},{"type":"comment","comment":"Boris, any updates on this?","author":"durin42","id":"28389","dateCreated":"1513015895","dateModified":"1513015895"},{"type":"comment","comment":"Do you think `DETACH_PROCESS` should remain specified? I believe that's\nthe source of the problem, but I can't say it's unnecessary with my little Windows-fu.\n\nWhat I know is our `win32.spawndetached()` sets only `CREATE_NO_WINDOW`.","author":"yuja","id":"26008","dateCreated":"1511959228","dateModified":"1511959228"},{"type":"update","diffId":"3914","author":"lothiraldan","id":"25821","dateCreated":"1511901396","dateModified":"1511901396"},{"type":"comment","comment":">>> `shell=False` also disables envar expansion, and I guess there would\n>>> be lots of trivial behavior changes.\n>> \n>> I'm not sure to understand what you mean, I tried this configuration:\n>> \n>> ```\n>> [logtoprocess]\n>> commandfinish = %HOMEPATH%\/script.pl\n>> ```\n>> \n>> And `%HOMEPATH` seemed to be expanded.\n> \n> Maybe it's resolved while looking up the executable path. I suspect\n> `script.pl %HOMEPATH%` wouldn't work.\n\nWith further testing, I can confirm that `script.pl %HOMEPATH%` is not working on Windows with `shell=False`. I don't know an equivalent of `shlex.split` for Windows, so I guess that documenting it would be enough as we don't force `shell=False`.\n\nI will document it.","author":"lothiraldan","id":"25813","dateCreated":"1511896687","dateModified":"1511896687"},{"type":"comment","comment":">>! In D1426#23803, @lothiraldan wrote:\n>>>! In D1426#23797, @yuja wrote:\n>> Can't we use `openflags=CREATE_NO_WINDOW` instead?\n>> \n>> https:\/\/stackoverflow.com\/a\/7006424\n> \n> I tried lots of things including openflags=CREATE_NO_WINDOW that didn't seem to work, unfortunately.\n\nRight. cmd.exe does allocates new console because the process is \"DETACHED\" from\nthe parent's console. I have no idea if `DETACH_PROCESS` is what we wanted.\n\n>> `shell=False` also disables envar expansion, and I guess there would\n>> be lots of trivial behavior changes.\n> \n> I'm not sure to understand what you mean, I tried this configuration:\n> \n> ```\n> [logtoprocess]\n> commandfinish = %HOMEPATH%\/script.pl\n> ```\n> \n> And `%HOMEPATH` seemed to be expanded.\n\nMaybe it's resolved while looking up the executable path. I suspect\n`script.pl %HOMEPATH%` wouldn't work.","author":"yuja","id":"23804","dateCreated":"1510842452","dateModified":"1510842452"},{"type":"comment","comment":">>! In D1426#23797, @yuja wrote:\n> Can't we use `openflags=CREATE_NO_WINDOW` instead?\n> \n> https:\/\/stackoverflow.com\/a\/7006424\n\nI tried lots of things including openflags=CREATE_NO_WINDOW that didn't seem to work, unfortunately.\n\n> \n> `shell=False` also disables envar expansion, and I guess there would\n> be lots of trivial behavior changes.\n\nI'm not sure to understand what you mean, I tried this configuration:\n\n```\n[logtoprocess]\ncommandfinish = %HOMEPATH%\/script.pl\n```\n\nAnd `%HOMEPATH` seemed to be expanded.","author":"lothiraldan","id":"23803","dateCreated":"1510839765","dateModified":"1510839765"},{"type":"comment","comment":"Can't we use `openflags=CREATE_NO_WINDOW` instead?\n\nhttps:\/\/stackoverflow.com\/a\/7006424\n\n`shell=False` also disables envar expansion, and I guess there would\nbe lots of trivial behavior changes.","author":"yuja","id":"23798","dateCreated":"1510838357","dateModified":"1510838357"},{"type":"inline","comment":"Right, I was not aware of this limitation.\n\nIf a company would deploy the same configuration for both Windows and Linux, we might have `shell=False` on Linux system but it would work well if the script doesn't take any argument or with an intermediary script that injects all the arguments, I will document it in the commit message and extensions documentation.","replyTo":"23570","isNewFile":"1","line":"65","lineLength":"0","path":"hgext\/logtoprocess.py","diffId":"3528","author":"lothiraldan","id":"23793","dateCreated":"1510823364","dateModified":"1510823364"},{"type":"inline","comment":"On unix systems `shell=False` means the command provided cannot take any arguments, as it's the shell that does the argument splitting.  I am guessing that this is OK, as we'll just never configure `logtoprocess.shell` as False on unix systems.  Is that right?","replyTo":null,"isNewFile":"1","line":"65","lineLength":"0","path":"hgext\/logtoprocess.py","diffId":"3528","author":"mbthomas","id":"23570","dateCreated":"1510766283","dateModified":"1510766283"},{"type":"accept","author":"mbthomas","id":"23569","dateCreated":"1510766283","dateModified":"1510766283"},{"type":"update","diffId":"3528","author":"lothiraldan","id":"23538","dateCreated":"1510760656","dateModified":"1510760656"}],"dateCreated":"1510760656","dateModified":"1513333033","status":"Needs Review"},{"id":"1700","callsign":"HG","title":"logtoprocess: pass ui in runshellcommand","author":"lothiraldan","summary":"We are gonna read a configuration specific to Windows, instead of adding a\nuseless arguments for runshellcommand on linux, pass the ui object to read it\ndirectly.","testPlan":"","lineCount":"6","dependsOn":[],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"4467","author":"lothiraldan","id":"29023","dateCreated":"1513332655","dateModified":"1513332655"}],"dateCreated":"1513332655","dateModified":"1513332916","status":"Needs Review"},{"id":"1428","callsign":"HG","title":"logtoprocess: connect all fds to \/dev\/null to avoid bad interaction with pager","author":"lothiraldan","summary":"We detected that pager is waiting for log-to-process script to finish, which\nis annoying when adding a script on commandfinish that does an HTTP push.\n\nThere seems to be no workaround on the script side and it will make the\nbehavior on Linux\/MacOS closer to the Windows behavior.\n\nThe drawback is that it makes the related tests more flaky as log-to-process\noutputs are now really asynchronous.\n\nIf it's considered a BC change, another option would be to add a config option\nfor this new behavior. I personally think that the different behavior between\nWindows and Linux is confusing and that it's a bug I would be fine with a new\nconfig option.","testPlan":"","lineCount":"15","dependsOn":["1427"],"reviewers":["yuja"],"ccs":["yuja","mercurial-devel"],"actions":[{"type":"update","diffId":"4469","author":"lothiraldan","id":"29038","dateCreated":"1513332668","dateModified":"1513332668"},{"type":"update","diffId":"3917","author":"lothiraldan","id":"25833","dateCreated":"1511901404","dateModified":"1511901404"},{"type":"update","diffId":"3589","author":"lothiraldan","id":"23965","dateCreated":"1510907108","dateModified":"1510907108"},{"type":"inline","comment":"Thx for the catch, I will make the change.","replyTo":"23801","isNewFile":"1","line":"89","lineLength":"0","path":"hgext\/logtoprocess.py","diffId":"3530","author":"lothiraldan","id":"23963","dateCreated":"1510905810","dateModified":"1510905810"},{"type":"inline","comment":"stdout and stderr should be writable. We can simply pass a file\ndescriptor open with O_RDWR.\n\n```\nnullfd = os.open(os.devnull, os.O_RDWR)\nPopen(stdin=nullfd, stdout=nullfd, stderr=nullfd)\n```\nand `os.close(nullfd)` though it doesn't mater since the process\nis terminated with `_exit()`.","replyTo":null,"isNewFile":"1","line":"89","lineLength":"0","path":"hgext\/logtoprocess.py","diffId":"3530","author":"yuja","id":"23801","dateCreated":"1510839259","dateModified":"1510839259"},{"type":"reject","author":"yuja","id":"23800","dateCreated":"1510839259","dateModified":"1510839259"},{"type":"update","diffId":"3530","author":"lothiraldan","id":"23558","dateCreated":"1510760666","dateModified":"1510760666"}],"dateCreated":"1510760666","dateModified":"1513332813","status":"Needs Review"},{"id":"1520","callsign":"FBHGX","title":"clindex: integrate Rust childmap","author":"quark","summary":"Like nodemap, maintain the Rust childmap in clindex. The childmap is pretty\ncompact - only 10% the size of changelog.i in fbsource.","testPlan":"","lineCount":"92","dependsOn":["1519"],"reviewers":[],"ccs":[],"actions":[{"type":"update","diffId":"4457","author":"quark","id":"28984","dateCreated":"1513304503","dateModified":"1513304503"},{"type":"update","diffId":"3882","author":"quark","id":"25618","dateCreated":"1511807454","dateModified":"1511807454"}],"dateCreated":"1511807454","dateModified":"1513307719","status":"Needs Review"},{"id":"1519","callsign":"FBHGX","title":"childmap: add a map for efficient children query","author":"quark","summary":" ","testPlan":"","lineCount":"509","dependsOn":["1518"],"reviewers":[],"ccs":[],"actions":[{"type":"update","diffId":"4462","author":"quark","id":"28998","dateCreated":"1513307718","dateModified":"1513307718"},{"type":"update","diffId":"4456","author":"quark","id":"28983","dateCreated":"1513304501","dateModified":"1513304501"},{"type":"update","diffId":"3881","author":"quark","id":"25610","dateCreated":"1511807452","dateModified":"1511807452"}],"dateCreated":"1511807452","dateModified":"1513307718","status":"Needs Review"},{"id":"1534","callsign":"HG","title":"logtoprocess: add an option to parse script command","author":"lothiraldan","summary":"With `shell=False`, the logtoprocess script will not be parsed by a shell and\nwe won't be able to pass arguments in logtoprocess configuration. Add a new\noption `logtoprocess.parsescript` that calls `shlex.split` on the script so we\ncan pass arguments. As `shlex.split` is only designed for POSIX system, this\noption have no impact on Windows.","testPlan":"","lineCount":"16","dependsOn":["1426"],"reviewers":[],"ccs":["mercurial-devel"],"actions":[{"type":"update","diffId":"3915","author":"lothiraldan","id":"25823","dateCreated":"1511901399","dateModified":"1511901399"}],"dateCreated":"1511901399","dateModified":"1513269543","status":"Needs Review"},{"id":"1330","callsign":"FBHGX","title":"hgext3rd: add hg continue and hg abort","author":"phillco","summary":"I hate having to remember and type which command I need to abort or continue.\nWhy make the user do that work? Life is short, and we already know what\nstatefiles exist and how to deal with them.\n\nWe could even have it loop until no statefiles are left, to deal with issues\nwhere both a rebase state and a merge state are left.","testPlan":"","lineCount":"63","dependsOn":[],"reviewers":["quark","durham"],"ccs":["durham","quark"],"actions":[{"type":"inline","comment":"Yeh, that's fair. It's a shame merge (the command) doesn't have a distinct statefile from merge.update (the function). Since we don't need to support the `hg merge` case at FB, I'll just have it detect the merge state file and print something in that case.","replyTo":"22386","isNewFile":"1","line":"68","lineLength":"0","path":"hgext3rd\/conflictinfo.py","diffId":"3348","author":"phillco","id":"22398","dateCreated":"1510247720","dateModified":"1510247720"},{"type":"inline","comment":"Can we print a message saying what we're continueing\/aborting?  Like \"continuing rebase (via hg rebase --continue) ...\" or \"aborting merge (via hg update --clean) ...\".  That way users know what action is being continued and knows what actually happening so they can figure out how to undo it if necessary.","replyTo":null,"isNewFile":"1","line":"34","lineLength":"0","path":"hgext3rd\/continue.py","diffId":"3348","author":"durham","id":"22387","dateCreated":"1510245420","dateModified":"1510245420"},{"type":"inline","comment":"Is merge\/state used in any case other than hg merge? I have this feeling it is?  If so, update --clean could lose user data and I don't think it's necessarily clear that 'hg abort' would do that.","replyTo":null,"isNewFile":"1","line":"68","lineLength":"0","path":"hgext3rd\/conflictinfo.py","diffId":"3348","author":"durham","id":"22386","dateCreated":"1510245420","dateModified":"1510245420"},{"type":"comment","comment":"Probably needs a test once it's not a POC","author":"durham","id":"22385","dateCreated":"1510245420","dateModified":"1510245420"},{"type":"reject","author":"durham","id":"22384","dateCreated":"1510245420","dateModified":"1510245420"},{"type":"inline","comment":"Add the `from __future__ import absolute_import` here.","replyTo":null,"isNewFile":"1","line":"7","lineLength":"0","path":"hgext3rd\/continue.py","diffId":"3348","author":"quark","id":"22268","dateCreated":"1510169291","dateModified":"1510169291"},{"type":"inline","comment":"Use `from . import conflicttinfo` instead.","replyTo":null,"isNewFile":"1","line":"20","lineLength":"0","path":"hgext3rd\/continue.py","diffId":"3348","author":"quark","id":"22267","dateCreated":"1510169291","dateModified":"1510169291"},{"type":"update","diffId":"3348","author":"phillco","id":"22256","dateCreated":"1510162410","dateModified":"1510162410"},{"type":"inline","comment":"It might be obviated by your suggestion. I think it's related to dispatching twice. I'll debug it more if we actually want to go this way.","replyTo":"22072","isNewFile":"1","line":"127","lineLength":"1","path":"hgext3rd\/undo.py","diffId":"3317","author":"phillco","id":"22075","dateCreated":"1510092299","dateModified":"1510092299"},{"type":"inline","comment":"That might work better. I'll try it.","replyTo":"22071","isNewFile":"1","line":"27","lineLength":"0","path":"hgext3rd\/continue.py","diffId":"3317","author":"phillco","id":"22074","dateCreated":"1510092299","dateModified":"1510092299"},{"type":"inline","comment":"Why is this change?","replyTo":null,"isNewFile":"1","line":"127","lineLength":"1","path":"hgext3rd\/undo.py","diffId":"3317","author":"quark","id":"22072","dateCreated":"1510092104","dateModified":"1510092104"},{"type":"inline","comment":"Instead of issuing a new dispatch request, you can lookup the command function using `cmdutil.findcmd` and execute that function directly.","replyTo":null,"isNewFile":"1","line":"27","lineLength":"0","path":"hgext3rd\/continue.py","diffId":"3317","author":"quark","id":"22071","dateCreated":"1510092104","dateModified":"1510092104"},{"type":"reject","author":"quark","id":"22070","dateCreated":"1510092104","dateModified":"1510092104"},{"type":"update","diffId":"3317","author":"phillco","id":"22034","dateCreated":"1510080495","dateModified":"1510080495"},{"type":"update","diffId":"3316","author":"phillco","id":"22033","dateCreated":"1510080441","dateModified":"1510080441"},{"type":"update","diffId":"3315","author":"phillco","id":"22029","dateCreated":"1510080374","dateModified":"1510080374"}],"dateCreated":"1510080374","dateModified":"1513209943","status":"Needs Revision"},{"id":"1665","callsign":"FBHGX","title":"perftweaks: pass a correct revgen to branchcache.update","author":"quark","summary":"The revgen parameter cannot be null. Otherwise the \"for i in None\" code path\nwill be executed if disablebranchcache is False while disablebranchcache2 is\nTrue.","testPlan":"","lineCount":"2","dependsOn":[],"reviewers":[],"ccs":[],"actions":[{"type":"update","diffId":"4373","author":"quark","id":"28483","dateCreated":"1513039161","dateModified":"1513039161"}],"dateCreated":"1513039161","dateModified":"1513039161","status":"Needs Review"},{"id":"1622","callsign":"HG","title":"[PoC] transaction: Use intbitset for implementing changes['phase']","author":"joerg.sonnenberger","summary":" ","testPlan":"","lineCount":"4","dependsOn":["1621"],"reviewers":["yuja"],"ccs":["yuja","quark","mercurial-devel"],"actions":[{"type":"comment","comment":">>! In D1622#27981, @joerg.sonnenberger wrote:\n> Like I said, this is primarily a proof of concept. I'm still playing with different approaches\n\nAh, okay. I've added [PoC] to the subject so this won't be queued by mistake.\n\nFWIW, we generally don't move codes to C unless there is significant improvement.\n\n","author":"yuja","id":"28278","dateCreated":"1512904090","dateModified":"1512904090"},{"type":"comment","comment":"Like I said, this is primarily a proof of concept. I'm still playing with different approaches and I don't think intbitset is a good fit for other reasons, but it worked the best from the various pre-existing implementations.","author":"joerg.sonnenberger","id":"27981","dateCreated":"1512832551","dateModified":"1512832551"},{"type":"comment","comment":"> Although this is definitely better for the clone case. intbitset does not seem to implement segmented bitsets so the memory usage could in theory be worse for other cases.\n\nPerhaps. And if we need to iterate a sparse bitset, the performance would\nbe way worse.\n","author":"yuja","id":"27979","dateCreated":"1512825512","dateModified":"1512825512"},{"type":"reject","author":"yuja","id":"27978","dateCreated":"1512825512","dateModified":"1512825512"},{"type":"inline","comment":"It's still worthwhile to have `set()` as a fallback. So the code runs in pure python or pypy environment","replyTo":null,"isNewFile":"1","line":"1283","lineLength":"0","path":"mercurial\/localrepo.py","diffId":"4279","author":"quark","id":"27920","dateCreated":"1512779474","dateModified":"1512779474"},{"type":"comment","comment":"Although this is definitely better for the clone case. intbitset does not seem to implement segmented bitsets so the memory usage could in theory be worse for other cases.","author":"quark","id":"27919","dateCreated":"1512779474","dateModified":"1512779474"},{"type":"update","diffId":"4279","author":"joerg.sonnenberger","id":"27903","dateCreated":"1512777992","dateModified":"1512777992"}],"dateCreated":"1512777992","dateModified":"1512904090","status":"Needs Revision"},{"id":"1618","callsign":"FBHGX","title":"infinitepush: write backup state file atomically","author":"stash","summary":"","testPlan":"Run infinitepush tests","lineCount":"2","dependsOn":[],"reviewers":["mbthomas","quark"],"ccs":["quark"],"actions":[{"type":"inline","comment":"`json.dump({...}, f)` could avoid creating the temporary string. Although it probably does not make a visible difference here.","replyTo":null,"isNewFile":"1","line":"842","lineLength":"0","path":"infinitepush\/backupcommands.py","diffId":"4255","author":"quark","id":"27946","dateCreated":"1512788509","dateModified":"1512788509"},{"type":"accept","author":"quark","id":"27945","dateCreated":"1512788509","dateModified":"1512788509"},{"type":"update","diffId":"4255","author":"stash","id":"27810","dateCreated":"1512751899","dateModified":"1512751899"}],"dateCreated":"1512751899","dateModified":"1512788509","status":"Accepted"},{"id":"1593","callsign":"FBHGX","title":"infinitepush: remove duplicated logic for storing changegroups","author":"pulkit","summary":"Before this patch, we had same code in parthandler for scratchbranchparttype and\nprocesspart where we override bundle2.processpart and add storing changegroup\nlogic if we get scratchbranchparttype.\n\nSince we have all this logic in parthandler, we can just go ahead and process\nthe scratchbranchparttype instead of adding it in processparts().","testPlan":"","lineCount":"24","dependsOn":[],"reviewers":["durham"],"ccs":[],"actions":[{"type":"inline","comment":"Ah, thanks for the explanation. I got confused by the exact same code and variable names.","replyTo":"27688","isNewFile":"0","line":"1125","lineLength":"0","path":"infinitepush\/__init__.py","diffId":"4121","author":"pulkit","id":"27689","dateCreated":"1512694616","dateModified":"1512694616"},{"type":"inline","comment":"This one is storing the entire bundler though.  So every part that was added to the bundler above is in the infinitepush stored bundle.  The standard scratchbranchparttype part handler would result in a bundle that only contained the changegroup part.  So the new code allows us to produce a bundle that contain parts like obsmarkers and stuff.","replyTo":"27659","isNewFile":"0","line":"1125","lineLength":"0","path":"infinitepush\/__init__.py","diffId":"4121","author":"durham","id":"27688","dateCreated":"1512693851","dateModified":"1512693851"},{"type":"inline","comment":"@durham I am referring to this part handler part. It has the same logic which I deleted above.","replyTo":null,"isNewFile":"0","line":"1254","lineLength":"0","path":"infinitepush\/__init__.py","diffId":"4121","author":"pulkit","id":"27660","dateCreated":"1512685256","dateModified":"1512685256"},{"type":"inline","comment":"There is part handler defined for that already which has exact same logic which I just deleted. The only difference I see is that, in this case, we are storing changegroup in the end, while that won't be executed in the end.","replyTo":"27630","isNewFile":"0","line":"1125","lineLength":"0","path":"infinitepush\/__init__.py","diffId":"4121","author":"pulkit","id":"27659","dateCreated":"1512685256","dateModified":"1512685256"},{"type":"inline","comment":"If we delete this code, when does this bundler get processed? I think the purpose of having this serialization logic is so that every other part in the bundle gets processed above first, and then we serialize it here.  Without this code here, then the bundler is just thrown away and all the parts that were added to it are lost, and the only serialization ends up being from the scratchbranchparttype processing, which doesn't handle the other parts.\n\nOr am I forgetting something?","replyTo":null,"isNewFile":"0","line":"1125","lineLength":"0","path":"infinitepush\/__init__.py","diffId":"4121","author":"durham","id":"27630","dateCreated":"1512683960","dateModified":"1512683960"},{"type":"reject","author":"durham","id":"27629","dateCreated":"1512683960","dateModified":"1512683960"},{"type":"comment","comment":"All the `test-infinitepush-*` skips for me except one, so I am not sure whether this breaks something or not. Any help on how can I run the test locally will be of great help.","author":"pulkit","id":"27072","dateCreated":"1512501221","dateModified":"1512501221"},{"type":"update","diffId":"4121","author":"pulkit","id":"27067","dateCreated":"1512501082","dateModified":"1512501082"}],"dateCreated":"1512501082","dateModified":"1512694616","status":"Needs Revision"},{"id":"1594","callsign":"FBHGX","title":"prune: only prompt uncommit when revs = ['.']","author":"quark","summary":"This is a follow-up of D1335.\n\n`uncommit` only works for the dot revision. Therefore the advice shouldn't\nbe visible when stripping other revisions.\n\nBesides, added missing i18n functions and added a test.","testPlan":"","lineCount":"40","dependsOn":[],"reviewers":["mitrandir"],"ccs":[],"actions":[{"type":"update","diffId":"4122","author":"quark","id":"27080","dateCreated":"1512504540","dateModified":"1512504540"}],"dateCreated":"1512504540","dateModified":"1512506757","status":"Needs Review"},{"id":"1542","callsign":"FBHGX","title":"fbamend: prevent amending only some changes with --rebase\/fixup","author":"tja","summary":"When calling hg amend --rebase <some files>,\na subsequent rebase fails if <some files> doesn't cover\nall the changes in the working copy.\n\nAs there is currently no code for shelving\/unshelving\nthe remaining changes, let's switch off this\ncombination of options for now, to avoid\ngetting people into failed rebase situations.\n\n--fixup can't work for the same reasons.","testPlan":"$ cd fb-hgext\/tests rt\n# Ran 232 tests, 21 skipped, 0 failed.","lineCount":"9","dependsOn":[],"reviewers":["quark"],"ccs":["quark"],"actions":[{"type":"comment","comment":"Probably also defend against:\n\n    --interactive\n    --include\n    --exclude\n\n","author":"quark","id":"26055","dateCreated":"1511984854","dateModified":"1511984871"},{"type":"accept","author":"quark","id":"26054","dateCreated":"1511984854","dateModified":"1511984854"},{"type":"update","diffId":"3961","author":"tja","id":"26014","dateCreated":"1511965164","dateModified":"1511965164"}],"dateCreated":"1511965164","dateModified":"1511984854","status":"Accepted"},{"id":"1312","callsign":"FBHGX","title":"packfiles: add hg debugpackstatus","author":"phillco","summary":"@akushner thought it'd be helpful to have a dedicated debug command to run if packfile unpleasantness is suspected, and I thought it'd be useful too.\n\nAlso added some missing metrics on remotefilelog stores.","testPlan":"","lineCount":"171","dependsOn":[],"reviewers":["singhsrb","durham"],"ccs":["quark","durham","akushner"],"actions":[{"type":"inline","comment":"Interesting, I didn't know you could chain them. I'll take the shorter way.","replyTo":"25994","isNewFile":"1","line":"385","lineLength":"2","path":"remotefilelog\/debugcommands.py","diffId":"3887","author":"phillco","id":"26005","dateCreated":"1511936991","dateModified":"1511936991"},{"type":"inline","comment":"Yeah, it doesn't matter today but I was looking ahead if we were to generalize this function \/ move it into a util. Then I'd expect it to behave the Ruby way too (or at least have a different name, like `filterfalsy`).","replyTo":"25993","isNewFile":"1","line":"377","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3887","author":"phillco","id":"26004","dateCreated":"1511936991","dateModified":"1511936991"},{"type":"inline","comment":"Maybe more common in Python:\n\n  [p for s in store.stores if s for p in getpaths(s)]\n\nIt's much shorter. Although I personally do not enjoy list comprehensive that much.","replyTo":null,"isNewFile":"1","line":"385","lineLength":"2","path":"remotefilelog\/debugcommands.py","diffId":"3887","author":"quark","id":"25994","dateCreated":"1511927497","dateModified":"1511927497"},{"type":"inline","comment":"Yes. Whether `0` or `[]` should be `False` is a separate topic (I personally prefer the Ruby\/Lua way - only `nil` and `false` are False). It seems you're using it to filter paths, which is fine I guess?","replyTo":"25992","isNewFile":"1","line":"377","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3887","author":"quark","id":"25993","dateCreated":"1511927497","dateModified":"1511927497"},{"type":"inline","comment":"@quark looks like that removes other falsy values too, like 0. But, I guess that is desirable?","replyTo":"25933","isNewFile":"1","line":"377","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3887","author":"phillco","id":"25992","dateCreated":"1511921930","dateModified":"1511921930"},{"type":"inline","comment":"clever, thx","replyTo":"25934","isNewFile":"1","line":"47","lineLength":"0","path":"remotefilelog\/basestore.py","diffId":"3887","author":"phillco","id":"25991","dateCreated":"1511921930","dateModified":"1511921930"},{"type":"inline","comment":"No need to split.","replyTo":"25657","isNewFile":"1","line":"202","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3887","author":"quark","id":"25935","dateCreated":"1511917343","dateModified":"1511917343"},{"type":"inline","comment":"Maybe just `count = len(list(self._listkeys())` since that's shorter and does not need placeholders","replyTo":null,"isNewFile":"1","line":"47","lineLength":"0","path":"remotefilelog\/basestore.py","diffId":"3887","author":"quark","id":"25934","dateCreated":"1511917343","dateModified":"1511917343"},{"type":"inline","comment":"Could you use `filter(None, list)`? It is more common in python.","replyTo":"25931","isNewFile":"1","line":"377","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3887","author":"quark","id":"25933","dateCreated":"1511917343","dateModified":"1511917343"},{"type":"inline","comment":"It's based on Ruby\/Underscore.js's name for the same function. Then again, who ever said Ruby was a paragon of naming?","replyTo":"25918","isNewFile":"1","line":"377","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3887","author":"phillco","id":"25931","dateCreated":"1511916796","dateModified":"1511916796"},{"type":"inline","comment":"I'd use the words \"File Data Store\" and \"File History Store\" in the user facing code.  content and metadata were old terms and should probably be replaced throughout the code.","replyTo":null,"isNewFile":"1","line":"416","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3887","author":"durham","id":"25919","dateCreated":"1511915533","dateModified":"1511915533"},{"type":"inline","comment":"I'd rename this 'removenone' or something.","replyTo":null,"isNewFile":"1","line":"377","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3887","author":"durham","id":"25918","dateCreated":"1511915533","dateModified":"1511915533"},{"type":"accept","author":"durham","id":"25917","dateCreated":"1511915533","dateModified":"1511915533"},{"type":"inline","comment":"Not really related, but I noticed that `debugdatapack` aborted too early if one of the paths had any errors. I could split this out if you think it's warranted.","replyTo":null,"isNewFile":"1","line":"202","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3887","author":"phillco","id":"25657","dateCreated":"1511813176","dateModified":"1511813176"},{"type":"comment","comment":"I flushed this out a bit since the last change, adding `getmetrics()` to `datapackstore` and printing the paths for remotefilelog stores, too.","author":"phillco","id":"25656","dateCreated":"1511813176","dateModified":"1511813176"},{"type":"update","diffId":"3887","author":"phillco","id":"25655","dateCreated":"1511812988","dateModified":"1511812988"},{"type":"update","diffId":"3886","author":"phillco","id":"25654","dateCreated":"1511812858","dateModified":"1511812858"},{"type":"plan-changes","author":"phillco","id":"23792","dateCreated":"1510814137","dateModified":"1510814137"},{"type":"update","diffId":"3564","author":"phillco","id":"23790","dateCreated":"1510813934","dateModified":"1510813934"},{"type":"inline","comment":"os.path.join(self._getrootpath(), fnhashhex[:2], fnhashhex[2:]) would be the more platform agnostic way of doing this.","replyTo":null,"isNewFile":"1","line":"58","lineLength":"1","path":"remotefilelog\/basestore.py","diffId":"3509","author":"durham","id":"23657","dateCreated":"1510790200","dateModified":"1510790200"},{"type":"comment","comment":"Needs a test","author":"durham","id":"23656","dateCreated":"1510790200","dateModified":"1510790200"},{"type":"reject","author":"durham","id":"23655","dateCreated":"1510790200","dateModified":"1510790200"},{"type":"inline","comment":"From our IRL chat, added `verbose=False`.","replyTo":"21913","isNewFile":"1","line":"58","lineLength":"0","path":"remotefilelog\/basestore.py","diffId":"3289","author":"phillco","id":"23471","dateCreated":"1510716288","dateModified":"1510716288"},{"type":"comment","comment":"Ready for another look. Please check my C.","author":"phillco","id":"23470","dateCreated":"1510716288","dateModified":"1510716288"},{"type":"update","diffId":"3509","author":"phillco","id":"23468","dateCreated":"1510716229","dateModified":"1510716229"},{"type":"update","diffId":"3507","author":"phillco","id":"23465","dateCreated":"1510715278","dateModified":"1510715278"},{"type":"update","diffId":"3506","author":"phillco","id":"23464","dateCreated":"1510715111","dateModified":"1510715111"},{"type":"comment","comment":".","author":"phillco","id":"22115","dateCreated":"1510119235","dateModified":"1510119235"},{"type":"plan-changes","author":"phillco","id":"22114","dateCreated":"1510119235","dateModified":"1510119235"},{"type":"update","diffId":"3331","author":"phillco","id":"22109","dateCreated":"1510119227","dateModified":"1510119227"},{"type":"inline","comment":"We may want a try\/catch around all of these metric functions that call _getfiles.  _getfiles could throw an exception if the directory isn't readable for some reason, and we don't want metric logging code to bring down the process.  Or we add the try\/catch to _getfiles or _listkeys and print a warning if the directory is inaccesible, then return an empty list.","replyTo":null,"isNewFile":"1","line":"45","lineLength":"0","path":"remotefilelog\/basestore.py","diffId":"3289","author":"durham","id":"21916","dateCreated":"1509995914","dateModified":"1509995914"},{"type":"inline","comment":"If we're just getting the file size, we can use _listkeys to get the information necessary to compute the file path to get the size.","replyTo":null,"isNewFile":"1","line":"54","lineLength":"0","path":"remotefilelog\/basestore.py","diffId":"3289","author":"durham","id":"21915","dateCreated":"1509995914","dateModified":"1509995914"},{"type":"inline","comment":"If we're just counting, we can use _listkeys() instead, so we don't have to resolve the filenames.","replyTo":null,"isNewFile":"1","line":"46","lineLength":"0","path":"remotefilelog\/basestore.py","diffId":"3289","author":"durham","id":"21914","dateCreated":"1509995914","dateModified":"1509995914"},{"type":"inline","comment":"Isn't getmetrics called at store creation time now?  By your previous diff.  If so, getnumfiles is expensive right?  How do we ensure that we don't accidentally call getnumfiles every time we want to log metrics.","replyTo":null,"isNewFile":"1","line":"58","lineLength":"0","path":"remotefilelog\/basestore.py","diffId":"3289","author":"durham","id":"21913","dateCreated":"1509995914","dateModified":"1509995914"},{"type":"inline","comment":"I'd write out the full name `debugpackstatus`.  `st` isn't a super common acronym in the code base.","replyTo":null,"isNewFile":"1","line":"271","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3289","author":"durham","id":"21912","dateCreated":"1509995914","dateModified":"1509995914"},{"type":"reject","author":"durham","id":"21911","dateCreated":"1509995914","dateModified":"1509995914"},{"type":"update","diffId":"3289","author":"phillco","id":"21864","dateCreated":"1509935138","dateModified":"1509935138"},{"type":"update","diffId":"3266","author":"phillco","id":"21700","dateCreated":"1509736600","dateModified":"1509736600"}],"dateCreated":"1509736600","dateModified":"1511936991","status":"Accepted"},{"id":"1509","callsign":"FBHGX","title":"sparse: make use of treemanifest for far faster path listing","author":"mjpieters","summary":"","testPlan":"","lineCount":"25","dependsOn":["1508"],"reviewers":["durham"],"ccs":["durham"],"actions":[{"type":"comment","comment":"Maybe add a test for this?  You could add a test-treemanifest-sparse.tm.  test-treemanifest-blame.t is a minimal test case you could copy to get started.","author":"durham","id":"25914","dateCreated":"1511914858","dateModified":"1511914858"},{"type":"accept","author":"durham","id":"25913","dateCreated":"1511914858","dateModified":"1511914858"},{"type":"update","diffId":"3816","author":"mjpieters","id":"25379","dateCreated":"1511541341","dateModified":"1511541341"}],"dateCreated":"1511541341","dateModified":"1511914858","status":"Accepted"},{"id":"1508","callsign":"FBHGX","title":"sparse: clean up path handling, making it platform agnostic","author":"mjpieters","summary":"There is no need to strip path separators from the normalised path; normpath\nwill never leave any.\n\nWe also ensure that the path prefix we test ends in a path separator, to avoid\nmatching on sibling paths that happen to share a prefix.","testPlan":"","lineCount":"20","dependsOn":[],"reviewers":["durham"],"ccs":[],"actions":[{"type":"accept","author":"durham","id":"25910","dateCreated":"1511914664","dateModified":"1511914664"},{"type":"update","diffId":"3815","author":"mjpieters","id":"25374","dateCreated":"1511541316","dateModified":"1511541316"}],"dateCreated":"1511541316","dateModified":"1511914664","status":"Accepted"},{"id":"1524","callsign":"FBHGX","title":"morestatus: drop the extension as it was moved to core","author":"pulkit","summary":"This patch drops the morestatus extension as the functionality was directly\nmoved to core as `hg status -v` or can be used by setting config\n`commands.status.verbose` to True.\n\nNote: The update part of the morestatus was not ported to core as that wraps\nupdate command to save the args and dropping the extensions results in\nfunctionality lose of showing conflict state of hg update. But I think that\ncan't happen at Fb as `hg update` aborts if the working directory is dirty as\nmentioned in tweakdefaults.py.","testPlan":"","lineCount":"441","dependsOn":[],"reviewers":["quark"],"ccs":["quark"],"actions":[{"type":"comment","comment":"The update part can happen if users pressed Ctrl+C during a long update. So we still want that feature.","author":"quark","id":"25871","dateCreated":"1511910092","dateModified":"1511910092"},{"type":"reject","author":"quark","id":"25870","dateCreated":"1511910092","dateModified":"1511910092"},{"type":"update","diffId":"3896","author":"pulkit","id":"25678","dateCreated":"1511819079","dateModified":"1511819079"}],"dateCreated":"1511819079","dateModified":"1511910092","status":"Needs Revision"},{"id":"1525","callsign":"FBHGX","title":"fold: support user template in json output","author":"pulkit","summary":"This patch syncs the JSON output of fold command with that of core Mercurial\nextensions. For reference: https:\/\/phab.mercurial-scm.org\/D1293","testPlan":"","lineCount":"8","dependsOn":["1524"],"reviewers":[],"ccs":[],"actions":[{"type":"update","diffId":"3897","author":"pulkit","id":"25683","dateCreated":"1511819085","dateModified":"1511819085"}],"dateCreated":"1511819085","dateModified":"1511819085","status":"Needs Review"},{"id":"1506","callsign":"HG","title":"dagop: handle IndexError when using wdir() in dag range","author":"swhitaker","summary":"Using wdir() in a dag range revset can crash Mercurial. For example:\n\n    hg status --rev '.^::wdir()\n\nrevlog.c reports an IndexError in this instance, but it isn't caught\nby the calling code. This change adds IndexError to the set of exception\ntypes the calling code catches. When an IndexError is caught, the code\nfalls back to calling the pure Python implementation of reachableroots,\nwhich fails gracefully.","testPlan":"$ hg status --rev '.::wdir()'\nabort: working directory revision cannot be specified","lineCount":"2","dependsOn":[],"reviewers":["mbthomas","yuja"],"ccs":["ryanmce","yuja","mercurial-devel"],"actions":[{"type":"comment","comment":">>! In D1506#25509, @ryanmce wrote:\n> but ideally, wouldn't this be supported?\n\nYeah, this case, `includepath=True`, won't be difficult to handle. If reachableroots() raised\nWdirUnsupportedError, drop wdir revision from `y` of `x::y`, add parents instead, rerun\n`x::(y - wdir() + parents(wdir()))`, and append wdir to the result.","author":"yuja","id":"25511","dateCreated":"1511704872","dateModified":"1511704872"},{"type":"comment","comment":"I understand that fixing the crash is a good first step, but ideally, wouldn't this be supported? Is there a reason it's particularly hard to support?","author":"ryanmce","id":"25510","dateCreated":"1511689355","dateModified":"1511689355"},{"type":"comment","comment":"I think it's better to fail fast instead of trying the slow path which is supposed\nto raise WdirUnsupported exception.\n\n> $ hg status --rev '.::wdir()'\n> abort: working directory revision cannot be specified\n\nCan you add test?","author":"yuja","id":"25487","dateCreated":"1511582067","dateModified":"1511582067"},{"type":"reject","author":"yuja","id":"25486","dateCreated":"1511582067","dateModified":"1511582067"},{"type":"accept","author":"mbthomas","id":"25372","dateCreated":"1511541089","dateModified":"1511541089"},{"type":"update","diffId":"3813","author":"swhitaker","id":"25358","dateCreated":"1511525215","dateModified":"1511525215"}],"dateCreated":"1511525215","dateModified":"1511704872","status":"Needs Revision"},{"id":"474","callsign":"FBHGX","title":"perftweaks: store rev to node mapping for preloading","author":"mbthomas","summary":"Change perftweaks to store the node-to-rev mapping for interesting\nrevisions, rather than just the rev numbers.\n\nPreviously this was used to prepopulate the changelog's node-to-rev\nlookup trie, however since we only stored the revision numbers this\nstill required looking at the revlog index contents in order to\nread out the node hash.\n\nWith these changes, the trie can be prepopulated without touching\nthe index.\n\nOn its own, this gives only a modest performance increase, however\nwhen combined with using mmap for the revlog index (coming in\nanother diff), it gives a noticable improvment on repos with very\nlarge revlogs, as we no longer need to touch large portions of the\nrevlog index on every invocation.\n\nWhen loading the cache data, we check that the rev numbers are\nin range for the revlog, and drop any entries that are out of\nrange.  The assumption is that a node hash won't be reassigned to\nanother rev index.  The other way round ought to be ok, as the\nold hash ought to be abandoned if it's no longer part of the\nrevlog.  If there's any more checking needed, please suggest in\nthe review comments.","testPlan":"Re-run unit tests.\nGenerate cache data on a large repo and verify indexes can be\npreloaded without touching the index contents.\nRun with mmap change to verify perf improvement.","lineCount":"64","dependsOn":[],"reviewers":["durham"],"ccs":["quark"],"actions":[{"type":"comment","comment":"This is no longer needed after D1472.","author":"quark","id":"25010","dateCreated":"1511320901","dateModified":"1511320901"},{"type":"inline","comment":"Is this better than `revstr, node = line.split(':')`?\n\nAlso, mercurial generally avoids using `_` for unused variables, because `_` is already the function name for the string internationalization function.  I usually use `x` or some such","replyTo":null,"isNewFile":"1","line":"223","lineLength":"0","path":"hgext3rd\/perftweaks.py","diffId":"1155","author":"durham","id":"8042","dateCreated":"1503593933","dateModified":"1503593933"},{"type":"comment","comment":"I think we want to be pretty conservative about cache invalidation initially.  If we get the map wrong, we could end up reading entirely incorrect data out of the changelog and associating it with the incorrect node.  A few ideas:\n\n- We could initially treat the cache as risky, and instead of using the data we could just confirm that the data is always correct, by comparing the getnode(rev) result with our stored node->rev. If our cache doesn't match the truth, print a ui.develwarn so we see it.  Once we have confidence, we can actually use the map result instead of calling getnode(rev).\n- We should invalidate the entire cache when `repo.destroyed()` is called.  That is the function called at the end of a strip, to let extensions perform invalidation like this.\n- Maybe we could only use the cached mappings for commits older than `tip - 10000`.  Any node that claims to be newer than tip-10000 we could do getnode(rev) and put the actual node value in the nodemap instead.  Since the The top of the changelog is where the most churn happens and where the most risk will be, this should help prevent most bad cache issues and shouldn't add much IO slow down (10k revs is only 620KB of .i file)","author":"durham","id":"8041","dateCreated":"1503593933","dateModified":"1503593933"},{"type":"reject","author":"durham","id":"8040","dateCreated":"1503593933","dateModified":"1503593933"},{"type":"update","diffId":"1155","author":"mbthomas","id":"7443","dateCreated":"1503406015","dateModified":"1503406015"}],"dateCreated":"1503406015","dateModified":"1511320901","status":"Needs Revision"},{"id":"1453","callsign":"FBHGX","title":"remotefilelog: never deadlock on writing to a pipe","author":"ikostia","summary":"Instead of guesstimating that 10000 (or any fixed number) steps\nalways fit in a buffer, we can just count the bytes as we write.\n\nDepends on D1436","testPlan":"- not tested yet","lineCount":"28","dependsOn":[],"reviewers":["durham"],"ccs":["durham","wlis"],"actions":[{"type":"inline","comment":"`i += 1`","replyTo":null,"isNewFile":"1","line":"238","lineLength":"0","path":"remotefilelog\/fileserverclient.py","diffId":"3663","author":"durham","id":"24394","dateCreated":"1511203049","dateModified":"1511203049"},{"type":"comment","comment":"Have you tested the performance impact of this change?  Previously we would issue 10k at a time, now if the buffer is 65k and if each entry is 100 bytes (40 byte hash + 60 byte path) then we're only doing 650 at a time.  You may need to set `remotefilelog.cacheprocess=` to force it to fallback to normal ssh fetching to execute this code path in bulk.","author":"durham","id":"24393","dateCreated":"1511203049","dateModified":"1511203049"},{"type":"reject","author":"durham","id":"24392","dateCreated":"1511203049","dateModified":"1511203049"},{"type":"comment","comment":"remove unnecessary condition","author":"ikostia","id":"24339","dateCreated":"1511198823","dateModified":"1511198823"},{"type":"update","diffId":"3663","author":"ikostia","id":"24338","dateCreated":"1511198823","dateModified":"1511198823"},{"type":"inline","comment":"Is there a way to get this value from the OS, in case someone has some custom setting for the default pipe size?","replyTo":null,"isNewFile":"1","line":"372","lineLength":"0","path":"remotefilelog\/fileserverclient.py","diffId":"3630","author":"wlis","id":"24303","dateCreated":"1511198212","dateModified":"1511198212"},{"type":"inline","comment":"Probably remove the \"True\"","replyTo":null,"isNewFile":"1","line":"230","lineLength":"0","path":"remotefilelog\/fileserverclient.py","diffId":"3630","author":"wlis","id":"24302","dateCreated":"1511198212","dateModified":"1511198212"},{"type":"update","diffId":"3630","author":"ikostia","id":"24226","dateCreated":"1511125192","dateModified":"1511125192"}],"dateCreated":"1511125192","dateModified":"1511203049","status":"Needs Revision"},{"id":"1423","callsign":"FBHGX","title":"debugdatapack: add more stats, and --summary","author":"phillco","summary":"- Use byte and comma formatters\n- In addition to each filename, print stats for each packfile, and for everything together (if numpackfiles > 1)\n- Add a `--summary` option which hides all the nodes and filename totals, which can be very noisy for huge packs, and only shows the packfile\/grand-total lines.\n\nExample output: https:\/\/phabricator.intern.facebook.com\/P58641463","testPlan":"","lineCount":"577","dependsOn":[],"reviewers":["durham"],"ccs":["durham"],"actions":[{"type":"inline","comment":"It's not obvious what the '?' means on its own.  Maybe write out \"missing blob sizes\" or something","replyTo":null,"isNewFile":"1","line":"57","lineLength":"0","path":"tests\/test-treemanifest-server.t","diffId":"3512","author":"durham","id":"23653","dateCreated":"1510789627","dateModified":"1510789627"},{"type":"inline","comment":"Same here for files vs something else.","replyTo":null,"isNewFile":"1","line":"312","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3512","author":"durham","id":"23652","dateCreated":"1510789627","dateModified":"1510789627"},{"type":"inline","comment":"Instead of \"unique files\" maybe say \"unique paths\", since it could also represent trees, and eventually other things.","replyTo":null,"isNewFile":"1","line":"304","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3512","author":"durham","id":"23651","dateCreated":"1510789627","dateModified":"1510789627"},{"type":"inline","comment":"I'd move the new total up one line, and the above newline down one.  So\n```\nw:\nNode          Delta Base    Delta Length  Blob Size\nbb6ccd5dceaa  000000000000  2 bytes       2 bytes\nTotal:                      2 bytes       2 bytes   (0.0% bigger)\n\nx:\nNode ....\n```\n\nThat way the separation between files is more visible.","replyTo":null,"isNewFile":"1","line":"160","lineLength":"0","path":"tests\/test-remotefilelog-bgprefetch.t","diffId":"3512","author":"durham","id":"23650","dateCreated":"1510789627","dateModified":"1510789627"},{"type":"inline","comment":"This whole section (lines 253-297) is becoming a bit of a mess.  We should refactor it to be something like:\n```\ndef getfileentries(entryiter)\n    lastfilename = None\n    entries = []\n    for filename, node, deltabase, deltalen in entryiter:\n        if lastfilename is not None and filename != lastfilename:\n            yield lastfilename, entries\n            entries = []\n        lastfilename = filename\n        entries.append((node, deltabase, deltalen))\n\n    if lastfilename is not None:\n        yield lastfilename, entries    \n...\nentryiter = dpack.iterentries()\nfor filename, entries in getfileentries(entryiter):\n   # Print header\n   for entry in entries:\n      # print entry\n   # print summary\n```\nSo we can untangle the filename != lastfilename logic and the post-loop logic to handle the last case.","replyTo":null,"isNewFile":"1","line":"297","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3512","author":"durham","id":"23649","dateCreated":"1510789627","dateModified":"1510789627"},{"type":"inline","comment":"I'd move this two lines up.  It's not related to metadata but it's right between the metadata reading and using.","replyTo":null,"isNewFile":"1","line":"281","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3512","author":"durham","id":"23648","dateCreated":"1510789627","dateModified":"1510789627"},{"type":"inline","comment":"We should probably move the contents of this loop to another function.  Not in this diff of course.","replyTo":null,"isNewFile":"1","line":"225","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3512","author":"durham","id":"23647","dateCreated":"1510789627","dateModified":"1510789627"},{"type":"inline","comment":"Debugdatapack is getting big.  I'd at least move this inner function out to be a top level function. To start to reduce complexity.","replyTo":null,"isNewFile":"1","line":"219","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3512","author":"durham","id":"23646","dateCreated":"1510789627","dateModified":"1510789627"},{"type":"inline","comment":"not quite alphabetical","replyTo":null,"isNewFile":"1","line":"9","lineLength":"0","path":"remotefilelog\/debugcommands.py","diffId":"3512","author":"durham","id":"23645","dateCreated":"1510789627","dateModified":"1510789627"},{"type":"reject","author":"durham","id":"23644","dateCreated":"1510789627","dateModified":"1510789627"},{"type":"update","diffId":"3512","author":"phillco","id":"23492","dateCreated":"1510730331","dateModified":"1510730331"}],"dateCreated":"1510730331","dateModified":"1510789627","status":"Needs Revision"},{"id":"1371","callsign":"HG","title":"global: remove redundant parenthesis","author":"indygreg","summary":"My code editor was spewing several warnings for redundant\nparenthesis. It supports easily fixing them. So I just did it\nglobally.\n\nWhat redundant parenthesis are left are in 3rd party code or\nare needed around string literals so our style checker doesn't\ncomplain about the lack of _().","testPlan":"","lineCount":"490","dependsOn":[],"reviewers":["yuja"],"ccs":["yuja","dlax","mercurial-devel"],"actions":[{"type":"comment","comment":"I used my editor's feature to fix violations. That's probably what introduced the whitespace changes. If only Mercurial conformed to PEP-8 style...","author":"indygreg","id":"22888","dateCreated":"1510591352","dateModified":"1510591352"},{"type":"inline","comment":"I don't like this sort of changes because here a tuple is a\ndata structure used thoroughly, not an arbitrary set of values\nto be returned.","replyTo":null,"isNewFile":"0","line":"298","lineLength":"0","path":"mercurial\/revsetlang.py","diffId":"3426","author":"yuja","id":"22867","dateCreated":"1510580851","dateModified":"1510580851"},{"type":"reject","author":"yuja","id":"22866","dateCreated":"1510580851","dateModified":"1510580851"},{"type":"inline","comment":"I'd keep this one as it makes it clearer that we return a generator. Arguably, that's nitpicking.","replyTo":null,"isNewFile":"1","line":"319","lineLength":"0","path":"mercurial\/bookmarks.py","diffId":"3426","author":"dlax","id":"22834","dateCreated":"1510561193","dateModified":"1510561193"},{"type":"inline","comment":"Spurious indentation changes here as well.","replyTo":null,"isNewFile":"1","line":"337","lineLength":"0","path":"hgext\/hgk.py","diffId":"3426","author":"dlax","id":"22833","dateCreated":"1510561193","dateModified":"1510561193"},{"type":"inline","comment":"Why this extra indentation?","replyTo":null,"isNewFile":"1","line":"79","lineLength":"0","path":"hgext\/hgk.py","diffId":"3426","author":"dlax","id":"22832","dateCreated":"1510561193","dateModified":"1510561193"},{"type":"comment","comment":"LGTM modulo a few nits. Nice cleanup.","author":"dlax","id":"22831","dateCreated":"1510561193","dateModified":"1510561193"},{"type":"update","diffId":"3426","author":"indygreg","id":"22733","dateCreated":"1510455062","dateModified":"1510455062"}],"dateCreated":"1510455062","dateModified":"1510591352","status":"Needs Revision"},{"id":"1361","callsign":"NRWHG","title":"exchange: set \"known\" property in generic override","author":"martinvonz","summary":" ","testPlan":"","lineCount":"17","dependsOn":[],"reviewers":["durin42"],"ccs":[],"actions":[{"type":"accept","author":"durin42","id":"22631","dateCreated":"1510350907","dateModified":"1510350907"},{"type":"update","diffId":"3403","author":"martinvonz","id":"22572","dateCreated":"1510342722","dateModified":"1510342722"}],"dateCreated":"1510342722","dateModified":"1510350907","status":"Accepted"},{"id":"1362","callsign":"NRWHG","title":"tracked: access narrowpats from repo instead of re-reading","author":"martinvonz","summary":"This is how we do it throughout the rest of the file.","testPlan":"","lineCount":"2","dependsOn":["1361"],"reviewers":["durin42"],"ccs":[],"actions":[{"type":"accept","author":"durin42","id":"22629","dateCreated":"1510350886","dateModified":"1510350886"},{"type":"update","diffId":"3404","author":"martinvonz","id":"22579","dateCreated":"1510342722","dateModified":"1510342722"}],"dateCreated":"1510342722","dateModified":"1510350886","status":"Accepted"},{"id":"1363","callsign":"NRWHG","title":"widening: stop passing old{in,ex}clude to _widen()","author":"martinvonz","summary":"The values are easily (and efficiently) accessible on the repo\ninstance, and not passing them will make simplify an upcoming patch.","testPlan":"","lineCount":"7","dependsOn":["1362"],"reviewers":["durin42"],"ccs":[],"actions":[{"type":"accept","author":"durin42","id":"22627","dateCreated":"1510350876","dateModified":"1510350876"},{"type":"update","diffId":"3405","author":"martinvonz","id":"22587","dateCreated":"1510342723","dateModified":"1510342723"}],"dateCreated":"1510342723","dateModified":"1510350876","status":"Accepted"},{"id":"1360","callsign":"NRWHG","title":"exchange: set {,old}{in,ex}cludepats in generic override","author":"martinvonz","summary":"With this patch, one can safely call exchange.pull() and it will pass\nall necessary parameters to the server.\n\nI think we may want to change the parameters so lack of\nold{in,ex}cludepats means that they're the same as the current\nones. Then widening would instead add new{in,ex}cludepats. But let's\ndiscuss that later.","testPlan":"","lineCount":"49","dependsOn":[],"reviewers":["durin42"],"ccs":[],"actions":[{"type":"accept","author":"durin42","id":"22625","dateCreated":"1510350770","dateModified":"1510350770"},{"type":"update","diffId":"3402","author":"martinvonz","id":"22569","dateCreated":"1510342711","dateModified":"1510342711"}],"dateCreated":"1510342711","dateModified":"1510350770","status":"Accepted"},{"id":"829","callsign":"FBHGX","title":"tweakdefaults: improve basic help text","author":"mbthomas","summary":"The default help ('hg help' or 'hg --help') shows only the built-in commands.\nWith all of our extensions, this is not a sensible list to show.  It also\nlists the enabled extensions, which is a long list that is not very useful\nto users.\n\nWe override the help command in the following circumstances:\n\n  * `hg`, or `hg non-existent-command`, to show a much smaller list of commands.\n  * `hg help`, or `hg --help`, to show the full list of relevant commands.\n\nOther invocations of `hg help` are not affected.","testPlan":"","lineCount":"88","dependsOn":[],"reviewers":["ryanmce"],"ccs":["quark"],"actions":[{"type":"comment","comment":"Going to think about this a bit more.  It may be worth considering what improvements we can make to upstream instead.","author":"mbthomas","id":"16244","dateCreated":"1507280551","dateModified":"1507280551"},{"type":"plan-changes","author":"mbthomas","id":"16243","dateCreated":"1507280551","dateModified":"1507280551"},{"type":"inline","comment":"If we still do this out-of-core, these could be config options.","replyTo":null,"isNewFile":"1","line":"113","lineLength":"9","path":"hgext3rd\/tweakdefaults.py","diffId":"2137","author":"quark","id":"14123","dateCreated":"1506624833","dateModified":"1506624833"},{"type":"comment","comment":"When thinking about it again, it seems desirable to show extension commands even without `-v`.\n\nI also think the upstream wants the alternative behavior too. Otherwise `rebase` not shown up in `hg help` just seems strange. It could be gated by a config option (ex. `commands.help.showextensioncommands`) to satisfy old users.\n\nWith `-v`, we can show commands marked as `EXPERIMENTAL`, `ADVANCED`, `DEPRECATED`. This is consistent with command flags.","author":"quark","id":"14122","dateCreated":"1506624833","dateModified":"1506624833"},{"type":"inline","comment":"This seems to be a copy-paste from core Mercurial to replace part of its function. Usually we avoid doing that unless the feature is urgent.\n\nConsider:\n\n - Get this feature done directly in upstream (I think `-v` is reasonable)\n\nOr:\n\n - Split upstream function so we can wrap it without code duplication. See 45e02cfad4bd for an example.","replyTo":null,"isNewFile":"1","line":"245","lineLength":"64","path":"hgext3rd\/tweakdefaults.py","diffId":"2137","author":"quark","id":"14121","dateCreated":"1506624349","dateModified":"1506624349"},{"type":"inline","comment":"We no longer support Python 2.6, the set literal syntax is preferred:\n\n  helpcommandsshort = {'add', ...}","replyTo":null,"isNewFile":"1","line":"113","lineLength":"0","path":"hgext3rd\/tweakdefaults.py","diffId":"2137","author":"quark","id":"14120","dateCreated":"1506624349","dateModified":"1506624349"},{"type":"update","diffId":"2137","author":"mbthomas","id":"14004","dateCreated":"1506601060","dateModified":"1506601060"},{"type":"comment","comment":"Feel free to bikeshed the list of commands that are presented.","author":"mbthomas","id":"14003","dateCreated":"1506600917","dateModified":"1506600917"},{"type":"update","diffId":"2136","author":"mbthomas","id":"13998","dateCreated":"1506600866","dateModified":"1506600866"}],"dateCreated":"1506600866","dateModified":"1507280551","status":"Changes Planned"}],"profiles":[{"userName":"jwatt","realName":"Jonathan Watt","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/4gwcogg7j7egn4mqjcwz\/PHID-FILE-y6yfth4iwcql6hrs4yeo\/alphanumeric_aleo-white_J.png-_335862-0%2C0%2C0%2C0.3.png"},{"userName":"abhyudaypratap","realName":"ABHYUDAY PRATAP SINGH","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/fhyoa4ylarfj4mjjsfnz\/PHID-FILE-gdiilvqk6dvaelqttevu\/alphanumeric_lato-white_A.png-_7867a3-0%2C0%2C0%2C0.png"},{"userName":"nspanti-logilab","realName":"Nicola Spanti","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/2szz77xfd5ejbn5ipcfr\/PHID-FILE-6ltvplbn3fcn7r5ogoqm\/alphanumeric_lato-white_N.png-_4e5632-0%2C0%2C0%2C0.3.png"},{"userName":"jeffpc","realName":"Jeff Sipek","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/tdb3ls7h66j7dh5upryy\/PHID-FILE-7uafys6btqxf3dd33pey\/alphanumeric_lato-white_J.png-_b1c8eb-0%2C0%2C0%2C0.3.png"},{"userName":"sheehan","realName":"Connor Sheehan","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/m4vwt2qskctqocukgsmu\/PHID-FILE-f4j7fdb5wpwxmj2pmi56\/profile"},{"userName":"idlsoft","realName":"Sandu Turcan","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/d5bwyakyn456l6k6dqyp\/PHID-FILE-l4g65sw3fcassfxteu7h\/profile"},{"userName":"sangeet259","realName":"Sangeet Kumar Mishra","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/w2t3qtjbkq6i7d4fiflb\/PHID-FILE-pjpjz4ab55pgexyec3j2\/alphanumeric_lato-dark_S.png-_4e5632-255%2C255%2C255%2C0.4.png"},{"userName":"Ivzhh","realName":"Sheng Mao","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/762wysb4ooe3ekoga43a\/PHID-FILE-zur24kxarfkspo26ir67\/profile"},{"userName":"khanchi97","realName":"Sushil khanchi","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/3pp56o77ancwuv65wlhu\/PHID-FILE-c2txpmsbutyxgjyde2zp\/profile"},{"userName":"kevincox","realName":"Kevin Cox","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/sbp5kmlbv6vamvjkmjy5\/PHID-FILE-knnjmppp4vpzcxpqmkwh\/profile"},{"userName":"dploch","realName":"Daniel Ploch","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/ociwxv6qntytpb6zdxls\/PHID-FILE-m4wajbifgsqubz5w6r4u\/profile"},{"userName":"tom.prince","realName":"Tom Prince","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/afbz4m4ohutpjetr3tql\/PHID-FILE-rblbnvxfl47fyw546ocf\/profile"},{"userName":"marco.leogrande","realName":"Marco Leogrande","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/tkvibdelmgssnv4ebov7\/PHID-FILE-tfjptcanefutqscooi75\/profile"},{"userName":"kaleyh","realName":"Kaley Zhihui Huang","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/clcmxxq7nzpdqhqeze35\/PHID-FILE-enqeoiscacbj3quvmu5v\/profile"},{"userName":"joerg.sonnenberger","realName":"Joerg Sonnenberger","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/r6zqtnrvelifbtppj4mu\/PHID-FILE-orm5fypdgkwptokuvhu5\/profile"},{"userName":"tja","realName":"Thomas Jacob","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/lwnnvur42bobvgmv7c54\/PHID-FILE-ztcsvphl242bhhrawoeh\/alphanumeric_aleo-white_T.png-_335862-0%2C0%2C0%2C0.png"},{"userName":"wlis","realName":"Wojciech Lis","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/qksokzpep5fnbwl4cxnv\/PHID-FILE-4snkiib4rpc643dsuvgs\/profile"},{"userName":"av6","realName":"Anton Shestakov","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/hbjxchhduli6jrz3wy53\/PHID-FILE-hixj32uzzbjuvjsax6e3\/profile"},{"userName":"foozy","realName":"Katsunori FUJIWARA","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/uxmk3ymbn5cplhrcc3ox\/PHID-FILE-eyxzcwqdufk55nuzrkui\/profile"},{"userName":"dlax","realName":"Denis Laxalde","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/llz4fftev4b7zckiqfvi\/PHID-FILE-d7t77nx6n56nzytddceb\/profile"},{"userName":"hollisb","realName":"Hollis Blanchard","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/iggfsbubv47fw5tdrrwe\/PHID-FILE-a57xt2zd72jp64xezauu\/alphanumeric_lato-white_H.png-_335862-255%2C255%2C255%2C0.7.png"},{"userName":"singhsrb","realName":"Saurabh Singh","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/bp7zbulgy4snhr3ou6z4\/PHID-FILE-dvujqtbp56tzyuttwvwp\/profile"},{"userName":"mbolin","realName":"Michael Bolin","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/a4aw73weeriuryfkpvbx\/PHID-FILE-nk3y36i25zsmgf7anqqh\/profile"},{"userName":"glandium","realName":"Mike Hommey","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/wexe6n4s2bpojsuxutj6\/PHID-FILE-angxhhuxuyx7rg7mniop\/alphanumeric_aleo-white_G.png-_6c8548-255%2C255%2C255%2C0.4.png"},{"userName":"mbthomas","realName":"Mark Thomas","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/ceuosb7vyo3urrgvwvnm\/PHID-FILE-ffbbyima2mbsxuqqisjf\/profile"},{"userName":"swhitaker","realName":"Simon Whitaker","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/rejzev6zgxqpmlisheps\/PHID-FILE-4mav5wnholgu7gerlvk6\/profile"},{"userName":"mjpieters","realName":"Martijn Pieters","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/7fln7u2rx2s3tivjclsm\/PHID-FILE-fv2jdrq3cueq5hjnosac\/profile"},{"userName":"rdamazio","realName":"Rodrigo Damazio Bovendorp","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/4a77tfxnpjbpt2ggycfj\/PHID-FILE-lsili2yioao6o54nuevs\/profile"},{"userName":"mharbison72","realName":"Matt Harbison","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/exavetnjpwt74m5ctwqc\/PHID-FILE-a6gptxixmw2meahkn2ad\/profile"},{"userName":"ikostia","realName":"Kostia Balytskyi","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/7zljfaajqc5fwpgggoxv\/PHID-FILE-6jwkzbmu4cf22o35mx3k\/profile"},{"userName":"lothiraldan","realName":"Boris Feld","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/m3bla7ufl3cjmnym5xat\/PHID-FILE-kigxoxqookrqrhpspjit\/profile"},{"userName":"stash","realName":"Stanislau Hlebik","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/wf3xb2kovxyqypqtp2rz\/PHID-FILE-akocypfziwimghtqh4iu\/alphanumeric_lato-dark_S.png-_9166aa-255%2C255%2C255%2C0.4.png"},{"userName":"indygreg","realName":"Gregory Szorc","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/nf3veituwivx24ctdwf6\/PHID-FILE-b2sr2l6owww6nd3m2kml\/profile"},{"userName":"dsp","realName":"David Soria Parra","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/xrjhk4aswt4nlwofuz2i\/PHID-FILE-rjdjfgxnzlwhx3iqgl3a\/profile"},{"userName":"spectral","realName":"Kyle Lippincott","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/afmemikabk26zacyrve3\/PHID-FILE-c7do5pqxf7rpec3jj3qs\/profile"},{"userName":"yuja","realName":"Yuya Nishihara","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/kyyyaiumhfkcjjtysphq\/PHID-FILE-iujngvd4rkw4tyyogag4\/profile"},{"userName":"ryanmce","realName":"Ryan McElroy","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/acfbd4z6nesfd76wnnj5\/PHID-FILE-k26xnkr6nvkc5l6jsscc\/profile"},{"userName":"pulkit","realName":"Pulkit Goyal","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/67mfdmcxxo6hghy7bxoq\/PHID-FILE-swcwmgf52vkrkmcqjnwd\/profile"},{"userName":"smf","realName":"Sean Farley","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/tbwekyw6pglxsobywzco\/PHID-FILE-hdxjtngw5tp2k5dbysdt\/profile"},{"userName":"durham","realName":"Durham Goode","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/tbr2anpbd33oiyqeijif\/PHID-FILE-g363nrrxd6paer5ugcpw\/profile"},{"userName":"jsgf","realName":"Jeremy Fitzhardinge","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/6ihycyjpnngswp5u3tum\/PHID-FILE-4x5bowxqnm7tdk27to2v\/profile"},{"userName":"martinvonz","realName":"Martin von Zweigbergk","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/llyfr65e22jrp77s6ufm\/PHID-FILE-fggemwj53prdzf5tnsir\/profile"},{"userName":"quark","realName":"Jun Wu","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/yhgthtf3uk3yamwaqfbv\/PHID-FILE-ntlso5gq5tmxn6qqmqhq\/profile"},{"userName":"phillco","realName":"Phil Cohen","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/s6v7x5h734dqbi4pkii6\/PHID-FILE-dcp3elsshux2rxyfb2g3\/profile"},{"userName":"krbullock","realName":"Kevin Bullock","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/cak2ma7v7iz7ac522on2\/PHID-FILE-achi52mww6qbb5dslmow\/profile"},{"userName":"durin42","realName":"Augie Fackler","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/fxdzcgfaiyjyat5yx6wn\/PHID-FILE-wab3a6a3bynn7dobtpoz\/profile"},{"userName":"admin","realName":"Administrator","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/ax2ohnusy6xhxcsictnf\/PHID-FILE-earziycgjj2v67gz3vq4\/alphanumeric_lato-white_A.png-_335862-0%2C0%2C0%2C0.3.png"},{"userName":"mercurial-devel","realName":"mercurial-devel@mercurial-scm.org","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/uxfm523cqrdxtqfl55gh\/PHID-FILE-igq3uy3gdohueaer3qxy\/alphanumeric_aleo-white_M.png-_335862-255%2C255%2C255%2C0.7.png"},{"userName":"mitrandir","realName":"Mateusz Jakub Kwapich","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/3og26nizhmuzz7vtuiob\/PHID-FILE-q3vx3ix5yiu5gs6umqvx\/profile"},{"userName":"akushner","realName":"Aaron Kushner","image":"https:\/\/phab.mercurial-scm.org\/file\/data\/z5rynd6nw66bxxba2mjl\/PHID-FILE-tgtgg2xrru4t4ezofo2k\/profile"}]}}