This is an archive of the discontinued Mercurial Phabricator instance.

Differential D6127

split: use the new movedirstate() we now have in scmutil
ClosedPublic

Authored by martinvonz on Mar 13 2019, 2:39 PM.

Details

Reviewers
pulkit
Group Reviewers
hg-reviewers
Commits
rHG42e2c7c52e1b: split: use the new movedirstate() we now have in scmutil
Summary

This avoids unnecessarily touching the working copy when splitting the
parent of the working copy. That also makes the test-removeemptydirs.t
case invalid, so we can just delete it.

Diff Detail

Repository
rHG Mercurial
Lint
Automatic diff as part of commit; lint not applicable.
Unit
Automatic diff as part of commit; unit tests not applicable.

Event Timeline

martinvonz created this revision.Mar 13 2019, 2:39 PM
pulkit added a subscriber: pulkit.Mar 15 2019, 9:55 AM
pulkit added inline comments.
tests/test-removeemptydirs.t
269

Do we have other tests covering such cases?

This test might be helpful in future for someone changing some behavior around the dirstate handling code.

martinvonz marked an inline comment as done.Mar 15 2019, 10:14 AM
martinvonz added inline comments.
tests/test-removeemptydirs.t
269

Not for split. It just doesn't happen anymore, because split no longer touches the working directory. We have other tests for removed directories in this file.

pulkit accepted this revision.Mar 16 2019, 7:03 PM
martinvonz marked an inline comment as done.Mar 16 2019, 7:57 PM
Closed by commit rHG42e2c7c52e1b: split: use the new movedirstate() we now have in scmutil (authored by martinvonz). · Explain Why
This revision was automatically updated to reflect the committed changes.

Revision Contents
Changeset List

PathPackages
Mhgext/split.py (11 lines)
Mtests/test-removeemptydirs.t (88 lines)

Diff 14525

hgext/split.py

bookmarks.activate(repo, bname) bookmarks.activate(repo, bname)
if torebase and top: if torebase and top:
dorebase(ui, repo, torebase, top) dorebase(ui, repo, torebase, top)
def dosplit(ui, repo, tr, ctx, opts): def dosplit(ui, repo, tr, ctx, opts):
committed = [] # [ctx] committed = [] # [ctx]
# Set working parent to ctx.p1(), and keep working copy as ctx's content # Set working parent to ctx.p1(), and keep working copy as ctx's content
# NOTE: if we can have "update without touching working copy" API, the if ctx.node() != repo.dirstate.p1():
# revert step could be cheaper. hg.clean(repo, ctx.node(), show_stats=False)
hg.clean(repo, ctx.p1().node(), show_stats=False) with repo.dirstate.parentchange():
parents = repo.changelog.parents(ctx.node()) scmutil.movedirstate(repo, ctx.p1())
ui.pushbuffer()
cmdutil.revert(ui, repo, ctx, parents)
ui.popbuffer() # discard "reverting ..." messages
# Any modified, added, removed, deleted result means split is incomplete # Any modified, added, removed, deleted result means split is incomplete
incomplete = lambda repo: any(repo.status()[:4]) incomplete = lambda repo: any(repo.status()[:4])
# Main split loop # Main split loop
while incomplete(repo): while incomplete(repo):
if committed: if committed:
header = (_('HG: Splitting %s. So far it has been split into:\n') header = (_('HG: Splitting %s. So far it has been split into:\n')

tests/test-removeemptydirs.t

$ hg log -T '{rev}:{node|short} {desc}\n' $ hg log -T '{rev}:{node|short} {desc}\n'
1:b9eddaa97cbc add foo 1:b9eddaa97cbc add foo
*** ***
add baz add baz
0:d17db4b0303a add bar 0:d17db4b0303a add bar
$ cd $TESTTMP $ cd $TESTTMP
Testing `hg split` being run from inside of a directory that was created in the
pulkitUnsubmitted
Done

Do we have other tests covering such cases?

This test might be helpful in future for someone changing some behavior around the dirstate handling code.

pulkit: Do we have other tests covering such cases? This test might be helpful in future for someone…
martinvonzAuthorUnsubmitted
Not Done

Not for split. It just doesn't happen anymore, because split no longer touches the working directory. We have other tests for removed directories in this file.

martinvonz: Not for split. It just doesn't happen anymore, because split no longer touches the working…
commit being split:
$ hg init hgsplit
$ cd hgsplit
$ cat >> .hg/hgrc << EOF
> [ui]
> interactive = 1
> [extensions]
> split =
> EOF
$ echo anchor > anchor.txt
$ hg ci -qAm anchor
Create a changeset with '/otherfile_in_root' and 'somedir/foo', then try to
split it.
$ echo otherfile > otherfile_in_root
$ mkdir somedir
$ cd somedir
$ echo hi > foo
$ hg ci -qAm split_me
(Note: need to make this file not in this directory, or else the bug doesn't
reproduce; we're using a separate file due to concerns of portability on
`echo -e`)
$ cat > ../split_commands << EOF
> n
> y
> y
> a
> EOF
The split succeeds on no-rmcwd platforms, which alters the rest of the tests
#if rmcwd
$ cat ../split_commands | hg split
current directory was removed
(consider changing to repo root: $TESTTMP/hgsplit)
diff --git a/otherfile_in_root b/otherfile_in_root
new file mode 100644
examine changes to 'otherfile_in_root'? [Ynesfdaq?] n
diff --git a/somedir/foo b/somedir/foo
new file mode 100644
examine changes to 'somedir/foo'? [Ynesfdaq?] y
@@ -0,0 +1,1 @@
+hi
record change 2/2 to 'somedir/foo'? [Ynesfdaq?] y
abort: $ENOENT$
[255]
#endif
Let's try that again without the rmdir
$ cd $TESTTMP/hgsplit/somedir
Show that the previous split didn't do anything
$ hg log -T '{rev}:{node|short} {desc}\n'
1:e26b22a4f0b7 split_me
0:7e53273730c0 anchor
$ hg status
? split_commands
Try again
$ cat ../split_commands | hg $NO_RM split
diff --git a/otherfile_in_root b/otherfile_in_root
new file mode 100644
examine changes to 'otherfile_in_root'? [Ynesfdaq?] n
diff --git a/somedir/foo b/somedir/foo
new file mode 100644
examine changes to 'somedir/foo'? [Ynesfdaq?] y
@@ -0,0 +1,1 @@
+hi
record change 2/2 to 'somedir/foo'? [Ynesfdaq?] y
created new head
diff --git a/otherfile_in_root b/otherfile_in_root
new file mode 100644
examine changes to 'otherfile_in_root'? [Ynesfdaq?] a
saved backup bundle to $TESTTMP/hgsplit/.hg/strip-backup/*-split.hg (glob)
Show that this split did something
$ hg log -T '{rev}:{node|short} {desc}\n'
2:a440f24fca4f split_me
1:c994f20276ab split_me
0:7e53273730c0 anchor
$ hg status
? split_commands