( )⚙ D1272 repack: sort orphan nodes by size

This is an archive of the discontinued Mercurial Phabricator instance.

Differential D1272

repack: sort orphan nodes by size
ClosedPublic

Authored by phillco on Oct 31 2017, 2:02 AM.
Tags
None
Subscribers
None

Details

Reviewers
durham
Group Reviewers
Restricted Project
Commits
rFBHGX7ac80b88cb99: repack: sort orphan nodes by size
Summary

Add repack.chainorphansbysize (default True).

When enabled, we take all orphaned nodes (nodes that are not part of a chain),
and put them into a new chain at the end, so we can get some minimal
compression out of them. Right now, they default to each being stored as
fulltexts, which is wasteful.

We sort the orphan chain by size, descending, to make the largest version
quickest to access, on the assumption that it is probably the newest. (This is
what Git does for its packed data, and it is a decent fallback if ancestry is
not available)

Example chain output, before:

A->B C D->E->F G H

After:

A->B D->E->F G->C->H
(assuming len(G)>=C=>H)

(I'm still adding a test case, but the code itself could be reviewed.)

Diff Detail

Repository
rFBHGX Facebook Mercurial Extensions
Lint
Lint Skipped
Unit
Unit Tests Skipped

Event Timeline

phillco created this revision.Oct 31 2017, 2:02 AM
Herald added a reviewer: Restricted Project. · View Herald TranscriptOct 31 2017, 2:02 AM
phillco edited the summary of this revision. (Show Details)Oct 31 2017, 2:06 AM
phillco edited the summary of this revision. (Show Details)
phillco edited the summary of this revision. (Show Details)
phillco edited the summary of this revision. (Show Details)Oct 31 2017, 2:08 AM
phillco retitled this revision from repack: sort orphan nodes by size to Add `repack.chainorphansbysize` (default True)..
phillco updated this revision to Diff 3174.
phillco added a reviewer: durham.Oct 31 2017, 2:11 AM
phillco retitled this revision from Add `repack.chainorphansbysize` (default True). to repack: sort orphans by size and chain together.
phillco edited the summary of this revision. (Show Details)
durham requested changes to this revision.Oct 31 2017, 6:42 PM

Any idea what the perf impact of this change is?

remotefilelog/repack.py
16

Missing trailing comma

521

A filter having side effects seems a bit funky. I'm not really convinced this is a cleaner pattern than just building a second list in a loop.

525

If we post-pone the listification until here, we could avoid creating two lists. i.e. the first generator would feed into the second directly.

579

Instead of depending on ordering (which I don't think guarantees that delta chain entries are directly next to each other), can we instead just keep track in the deltabase loop above of any commit that A) doesn't have a delta base, and B) isn't used as a delta base?

597

We probably should. This can result in flakey tests otherwise. Maybe just sort by node as a secondary key. That said, the python docs say sorted is guaranteed to be stable for things with the same key, assuming the input list has the same order each time. So maybe we just sort orphans by node first.

598

Instead of reversing the sorted output, you can pass reverse=True to sorted.

602

Maybe put this behind a 'if self.debugflag:' check. This loop can be a bit of a critical path, and doing a bunch of unnecessary string formats could add up.

613

This makes the function quite larger. Could we refactor this to be a separate function?

This revision now requires changes to proceed.Oct 31 2017, 6:42 PM
phillco retitled this revision from repack: sort orphans by size and chain together to repack: sort orphan nodes by size.Nov 6 2017, 12:51 PM
phillco updated this revision to Diff 3291.
phillco updated this revision to Diff 3292.
phillco planned changes to this revision.Nov 6 2017, 12:52 PM
phillco updated this revision to Diff 3307.Nov 7 2017, 11:55 AM
phillco updated this revision to Diff 3308.Nov 7 2017, 11:58 AM
durham requested changes to this revision.Nov 7 2017, 1:27 PM

Looks like this is still in progress? Putting back on your queue

This revision now requires changes to proceed.Nov 7 2017, 1:27 PM
phillco added a comment.Nov 7 2017, 6:41 PM

Yes, sorry about that. Submitted by mistake.

phillco marked 7 inline comments as done.Nov 10 2017, 1:00 AM
phillco edited the summary of this revision. (Show Details)
phillco updated this revision to Diff 3386.
phillco updated this revision to Diff 3387.Nov 10 2017, 1:09 AM
phillco added inline comments.Nov 10 2017, 1:10 AM
tests/test-remotefilelog-lfs.t
238–239 ↗(On Diff #3387)

Should probably add a hg debugdatapack $TESTTMP/hgcache/master/packs/*.datapack here on @ so we can see the type of change happening.

durham requested changes to this revision.Nov 14 2017, 11:57 AM

How much does this affect performance? Requesting changes for that information and for the potential KeyError I commented on.

remotefilelog/repack.py
458

Indentation isn't quite right.

538

Since this is a hotpath, we might want to avoid the double dictionary lookup. Might be better to do:

deltatuple = deltabases.get(node)
if deltatuple is None:
   nobase.add(node)
else:
  deltabase, chainlen = deltatuple
540–542

If node is not in deltabases, as the condition above checks, won't this line throw a KeyError?

This revision now requires changes to proceed.Nov 14 2017, 11:57 AM
phillco added inline comments.Nov 14 2017, 12:26 PM
remotefilelog/repack.py
540–542

No, because deltabases is a defaultdict. That's why we have to check first.

phillco added inline comments.Nov 14 2017, 12:29 PM
remotefilelog/repack.py
540–542

Although, now that I think about it, we can just assign the default here and make it a regular dict, which is probably clearer.

phillco added a comment.Nov 14 2017, 9:45 PM

@durham locally, I found it added 5s to a 70s full repack.

phillco updated this revision to Diff 3504.Nov 14 2017, 9:48 PM
phillco marked an inline comment as done.Nov 14 2017, 9:48 PM
phillco marked 5 inline comments as done.Nov 15 2017, 12:51 PM
phillco updated this revision to Diff 3531.
phillco added a comment.Nov 15 2017, 12:52 PM

(done)

remotefilelog/repack.py
538

Done. The only hitch is we have to write the nullid deltabase here too, otherwise it will throw a KeyError on one of the later loops.

durham accepted this revision.Nov 15 2017, 6:21 PM
durham added inline comments.
tests/test-remotefilelog-repack-fast.t
265

Kinda weird that the delta length is 70 and the blob size is 8 here? When it's not actually a delta? In the old version of the lines above, delta length and blob size matched when delta base was 0

Not related to this diff, but strange.

This revision is now accepted and ready to land.Nov 15 2017, 6:21 PM
phillco added inline comments.Nov 16 2017, 1:22 AM
tests/test-remotefilelog-repack-fast.t
265

Yeah, that is. I'll take a look if I can remember.

Revision Contents
Changeset List

Diff 3386

remotefilelog/repack.py

from __future__ import absolute_import from __future__ import absolute_import
import os import os
from collections import defaultdict from collections import defaultdict
from hgext3rd.extutil import runshellcommand from hgext3rd.extutil import runshellcommand
from mercurial import ( from mercurial import (
error, error,
extensions, extensions,
mdiff, mdiff,
policy, policy,
scmutil, scmutil,
util, util,
) )
from mercurial.node import nullid from mercurial.node import (
nullid,
short,
durhamUnsubmitted
Done

Missing trailing comma

durham: Missing trailing comma
)
from mercurial.i18n import _ from mercurial.i18n import _
from . import ( from . import (
constants, constants,
contentstore, contentstore,
datapack, datapack,
historypack, historypack,
metadatastore, metadatastore,
shallowutil, shallowutil,
# Run repack # Run repack
self.repackdata(ledger, targetdata) self.repackdata(ledger, targetdata)
self.repackhistory(ledger, targethistory) self.repackhistory(ledger, targethistory)
# Call cleanup on each source # Call cleanup on each source
for source in ledger.sources: for source in ledger.sources:
source.cleanup(ledger) source.cleanup(ledger)
def _chainorphans(self, ui, filename, nodes, orphans, deltabases):
"""Reorderes ``orphans`` into a single chain inside ``nodes`` and
``deltabases``.
We often have orphan entries (nodes without a base that aren't
referenced by other nodes -- i.e., part of a chain) due to gaps in
history. Rather than store them as individual fulltexts, we prefer to
insert them as one chain sorted by size.
"""
if not orphans:
return nodes
def getsize(node, default=0):
meta = self.data.getmeta(filename, node)
if constants.METAKEYSIZE in meta:
return meta[constants.METAKEYSIZE]
else:
return default
# Sort orphans by size; biggest first is preferred, since it's more
# likely to be the newest version assuming files grow over time.
# (Sort by node first to ensure the sort is stable.)
orphans = sorted(orphans)
orphans = list(sorted(orphans, key=getsize, reverse=True))
if ui.debugflag:
ui.debug("%s: orphan chain: %s\n" % (filename,
", ".join([short(s) for s in orphans])))
durhamUnsubmitted
Done

Indentation isn't quite right.

durham: Indentation isn't quite right.
# Create one contiguous chain and reassign deltabases.
for i, node in enumerate(orphans):
if i == 0:
deltabases[node] = (nullid, 0)
else:
parent = orphans[i - 1]
deltabases[node] = (parent, deltabases[parent][1] + 1)
nodes = filter(lambda node: node not in orphans, nodes)
nodes += orphans
return nodes
def repackdata(self, ledger, target): def repackdata(self, ledger, target):
ui = self.repo.ui ui = self.repo.ui
maxchainlen = ui.configint('packs', 'maxchainlen', 1000) maxchainlen = ui.configint('packs', 'maxchainlen', 1000)
byfile = {} byfile = {}
for entry in ledger.entries.itervalues(): for entry in ledger.entries.itervalues():
if entry.datasource: if entry.datasource:
byfile.setdefault(entry.filename, {})[entry.node] = entry byfile.setdefault(entry.filename, {})[entry.node] = entry
# Order the nodes children first, so we can produce reverse deltas # Order the nodes children first, so we can produce reverse deltas
orderednodes = list(reversed(self._toposort(ancestors))) orderednodes = list(reversed(self._toposort(ancestors)))
if len(nohistory) > 0: if len(nohistory) > 0:
ui.debug('repackdata: %d nodes without history\n' % ui.debug('repackdata: %d nodes without history\n' %
len(nohistory)) len(nohistory))
orderednodes.extend(sorted(nohistory)) orderednodes.extend(sorted(nohistory))
# Compute deltas and write to the pack # Filter orderednodes to just the nodes we want to serialize (it
deltabases = defaultdict(lambda: (nullid, 0)) # currently also has the edge nodes' ancestors).
nodes = set(nodes) orderednodes = filter(lambda node: node in nodes, orderednodes)
for i, node in enumerate(orderednodes):
# orderednodes is all ancestors, but we only want to serialize
# the files we have.
if node not in nodes:
continue
# Garbage collect old nodes:
if self.garbagecollect: if self.garbagecollect:
# If the node is old and is not in the keepset neworderednodes = []
# we skip it and mark as garbage collected for node in orderednodes:
# If the node is old and is not in the keepset, we skip it,
# and mark as garbage collected
if ((filename, node) not in self.keepkeys and if ((filename, node) not in self.keepkeys and
self.isold(self.repo, filename, node)): self.isold(self.repo, filename, node)):
durhamUnsubmitted
Done

A filter having side effects seems a bit funky. I'm not really convinced this is a cleaner pattern than just building a second list in a loop.

durham: A filter having side effects seems a bit funky. I'm not really convinced this is a cleaner…
entries[node].gced = True entries[node].gced = True
continue continue
neworderednodes.append(node)
orderednodes = neworderednodes
durhamUnsubmitted
Done

If we post-pone the listification until here, we could avoid creating two lists. i.e. the first generator would feed into the second directly.

durham: If we post-pone the listification until here, we could avoid creating two lists. i.e. the…
# Compute delta bases for nodes:
deltabases = defaultdict(lambda: (nullid, 0))
nobase = set()
referenced = set()
nodes = set(nodes)
for i, node in enumerate(orderednodes):
ui.progress(_("processing nodes"), i, unit='nodes', ui.progress(_("processing nodes"), i, unit='nodes',
total=len(orderednodes)) total=len(orderednodes))
# Find delta base # Find delta base
# TODO: allow delta'ing against most recent descendant instead # TODO: allow delta'ing against most recent descendant instead
# of immediate child # of immediate child
if node not in deltabases:
durhamUnsubmitted
Done

Since this is a hotpath, we might want to avoid the double dictionary lookup. Might be better to do:

deltatuple = deltabases.get(node)
if deltatuple is None:
   nobase.add(node)
else:
  deltabase, chainlen = deltatuple
durham: Since this is a hotpath, we might want to avoid the double dictionary lookup. Might be better…
phillcoAuthorUnsubmitted
Not Done

Done. The only hitch is we have to write the nullid deltabase here too, otherwise it will throw a KeyError on one of the later loops.

phillco: Done. The only hitch is we have to write the `nullid` deltabase here too, otherwise it will…
nobase.add(node)
deltabase, chainlen = deltabases[node] deltabase, chainlen = deltabases[node]
if deltabase != nullid:
referenced.add(deltabase)
durhamUnsubmitted
Done

If node is not in deltabases, as the condition above checks, won't this line throw a KeyError?

durham: If node is not in deltabases, as the condition above checks, won't this line throw a KeyError?
phillcoAuthorUnsubmitted
Done

No, because deltabases is a defaultdict. That's why we have to check first.

phillco: No, because deltabases is a `defaultdict`. That's why we have to check first.
phillcoAuthorUnsubmitted
Done

Although, now that I think about it, we can just assign the default here and make it a regular dict, which is probably clearer.

phillco: Although, now that I think about it, we can just assign the default here and make it a regular…
# Use available ancestor information to inform our delta choices # Use available ancestor information to inform our delta choices
ancestorinfo = ancestors.get(node) ancestorinfo = ancestors.get(node)
if ancestorinfo: if ancestorinfo:
p1, p2, linknode, copyfrom = ancestorinfo p1, p2, linknode, copyfrom = ancestorinfo
# The presence of copyfrom means we're at a point where the # The presence of copyfrom means we're at a point where the
# file was copied from elsewhere. So don't attempt to do any # file was copied from elsewhere. So don't attempt to do any
# deltas with the other file. # deltas with the other file.
if copyfrom: if copyfrom:
p1 = nullid p1 = nullid
if chainlen < maxchainlen: if chainlen < maxchainlen:
# Record this child as the delta base for its parents. # Record this child as the delta base for its parents.
# This may be non optimal, since the parents may have # This may be non optimal, since the parents may have
# many children, and this will only choose the last one. # many children, and this will only choose the last one.
# TODO: record all children and try all deltas to find # TODO: record all children and try all deltas to find
# best # best
if p1 != nullid: if p1 != nullid:
deltabases[p1] = (node, chainlen + 1) deltabases[p1] = (node, chainlen + 1)
if p2 != nullid: if p2 != nullid:
deltabases[p2] = (node, chainlen + 1) deltabases[p2] = (node, chainlen + 1)
if ui.configbool('repack', 'chainorphansbysize', True):
orphans = nobase - referenced
orderednodes = self._chainorphans(ui, filename, orderednodes,
orphans, deltabases)
# Compute deltas and write to the pack
for i, node in enumerate(orderednodes):
deltabase, chainlen = deltabases[node]
# Compute delta # Compute delta
# TODO: Optimize the deltachain fetching. Since we're # TODO: Optimize the deltachain fetching. Since we're
# iterating over the different version of the file, we may # iterating over the different version of the file, we may
# be fetching the same deltachain over and over again. # be fetching the same deltachain over and over again.
# TODO: reuse existing deltas if it matches our deltabase # TODO: reuse existing deltas if it matches our deltabase
if deltabase != nullid: if deltabase != nullid:
durhamUnsubmitted
Done

Instead of depending on ordering (which I don't think guarantees that delta chain entries are directly next to each other), can we instead just keep track in the deltabase loop above of any commit that A) doesn't have a delta base, and B) isn't used as a delta base?

durham: Instead of depending on ordering (which I don't think guarantees that delta chain entries are…
deltabasetext = self.data.get(filename, deltabase) deltabasetext = self.data.get(filename, deltabase)
original = self.data.get(filename, node) original = self.data.get(filename, node)
size = len(original) size = len(original)
delta = mdiff.textdiff(deltabasetext, original) delta = mdiff.textdiff(deltabasetext, original)
else: else:
delta = self.data.get(filename, node) delta = self.data.get(filename, node)
size = len(delta) size = len(delta)
# TODO: don't use the delta if it's larger than the fulltext # TODO: don't use the delta if it's larger than the fulltext
# TODO: don't use the delta if the chain is already long # TODO: don't use the delta if the chain is already long
meta = self.data.getmeta(filename, node) meta = self.data.getmeta(filename, node)
if constants.METAKEYSIZE not in meta: if constants.METAKEYSIZE not in meta:
meta[constants.METAKEYSIZE] = size meta[constants.METAKEYSIZE] = size
target.add(filename, node, deltabase, delta, meta) target.add(filename, node, deltabase, delta, meta)
entries[node].datarepacked = True entries[node].datarepacked = True
ui.progress(_("processing nodes"), None) ui.progress(_("processing nodes"), None)
durhamUnsubmitted
Done

We probably should. This can result in flakey tests otherwise. Maybe just sort by node as a secondary key. That said, the python docs say sorted is guaranteed to be stable for things with the same key, assuming the input list has the same order each time. So maybe we just sort orphans by node first.

durham: We probably should. This can result in flakey tests otherwise. Maybe just sort by node as a…
count += 1 count += 1
durhamUnsubmitted
Done

Instead of reversing the sorted output, you can pass reverse=True to sorted.

durham: Instead of reversing the sorted output, you can pass reverse=True to sorted.
ui.progress(_("repacking data"), None) ui.progress(_("repacking data"), None)
target.close(ledger=ledger) target.close(ledger=ledger)
durhamUnsubmitted
Done

Maybe put this behind a 'if self.debugflag:' check. This loop can be a bit of a critical path, and doing a bunch of unnecessary string formats could add up.

durham: Maybe put this behind a 'if self.debugflag:' check. This loop can be a bit of a critical path…
def repackhistory(self, ledger, target): def repackhistory(self, ledger, target):
ui = self.repo.ui ui = self.repo.ui
byfile = {} byfile = {}
for entry in ledger.entries.itervalues(): for entry in ledger.entries.itervalues():
if entry.historysource: if entry.historysource:
byfile.setdefault(entry.filename, {})[entry.node] = entry byfile.setdefault(entry.filename, {})[entry.node] = entry
count = 0 count = 0
for filename, entries in sorted(byfile.iteritems()): for filename, entries in sorted(byfile.iteritems()):
ancestors = {} ancestors = {}
durhamUnsubmitted
Done

This makes the function quite larger. Could we refactor this to be a separate function?

durham: This makes the function quite larger. Could we refactor this to be a separate function?
nodes = list(node for node in entries.iterkeys()) nodes = list(node for node in entries.iterkeys())
for node in nodes: for node in nodes:
if node in ancestors: if node in ancestors:
continue continue
ancestors.update(self.history.getancestors(filename, node, ancestors.update(self.history.getancestors(filename, node,
known=ancestors)) known=ancestors))

tests/test-remotefilelog-repack-fast.t

$ hg revert x $ hg revert x
$ hg commit -m 'add y back without metadata' $ hg commit -m 'add y back without metadata'
$ cd ../shallow $ cd ../shallow
$ hg pull -q $ hg pull -q
$ hg up -q tip $ hg up -q tip
2 files fetched over 2 fetches - (2 misses, 0.00% hit ratio) over * (glob) 2 files fetched over 2 fetches - (2 misses, 0.00% hit ratio) over * (glob)
$ hg repack $ hg repack
$ ls $TESTTMP/hgcache/master/packs $ ls $TESTTMP/hgcache/master/packs
e634f60d2a9539fc595b1f4db480c64556a396c7.dataidx e8fdf7ae22b772dcc291f905b9c6e5f381d28739.dataidx
e634f60d2a9539fc595b1f4db480c64556a396c7.datapack e8fdf7ae22b772dcc291f905b9c6e5f381d28739.datapack
ebbd7411e00456c0eec8d1150a77e2b3ef490f3f.histidx ebbd7411e00456c0eec8d1150a77e2b3ef490f3f.histidx
ebbd7411e00456c0eec8d1150a77e2b3ef490f3f.histpack ebbd7411e00456c0eec8d1150a77e2b3ef490f3f.histpack
$ hg debughistorypack $TESTTMP/hgcache/master/packs/*.histidx $ hg debughistorypack $TESTTMP/hgcache/master/packs/*.histidx
x x
Node P1 Node P2 Node Link Node Copy From Node P1 Node P2 Node Link Node Copy From
cd410a44d584 577959738234 000000000000 609547eda446 y cd410a44d584 577959738234 000000000000 609547eda446 y
1bb2e6237e03 d4a3ed9310e5 000000000000 0b03bbc9e1e7 1bb2e6237e03 d4a3ed9310e5 000000000000 0b03bbc9e1e7
$ rm -rf $TESTTMP/hgcache/master/packs $ rm -rf $TESTTMP/hgcache/master/packs
$ cp -R $TESTTMP/backuppacks $TESTTMP/hgcache/master/packs $ cp -R $TESTTMP/backuppacks $TESTTMP/hgcache/master/packs
# Test repacking datapack without history # Test repacking datapack without history
$ rm -rf $CACHEDIR/master/packs/*hist* $ rm -rf $CACHEDIR/master/packs/*hist*
$ hg repack $ hg repack
$ hg debugdatapack $TESTTMP/hgcache/master/packs/*.datapack $ hg debugdatapack $TESTTMP/hgcache/master/packs/*.datapack
$TESTTMP/hgcache/master/packs/055c02949317b8507cdb7aaf2e00cc00fd0c5716: $TESTTMP/hgcache/master/packs/a8d86ff8e1a11a77a85f5fea567f56a757583eda:
x: x:
Node Delta Base Delta Length Blob Size Node Delta Base Delta Length Blob Size
1bb2e6237e03 000000000000 8 8 1bb2e6237e03 000000000000 8 8
aee31534993a 000000000000 4 4 d4a3ed9310e5 1bb2e6237e03 12 6
d4a3ed9310e5 000000000000 6 6 aee31534993a d4a3ed9310e5 12 4
Total: 18 18 (0.0% bigger) Total: 32 18 (77.8% bigger)
y: y:
Node Delta Base Delta Length Blob Size Node Delta Base Delta Length Blob Size
577959738234 000000000000 70 8 577959738234 000000000000 70 8
durhamUnsubmitted
Not Done

Kinda weird that the delta length is 70 and the blob size is 8 here? When it's not actually a delta? In the old version of the lines above, delta length and blob size matched when delta base was 0

Not related to this diff, but strange.

durham: Kinda weird that the delta length is 70 and the blob size is 8 here? When it's not actually a…
phillcoAuthorUnsubmitted
Not Done

Yeah, that is. I'll take a look if I can remember.

phillco: Yeah, that is. I'll take a look if I can remember.
Total: 70 8 (775.0% bigger) Total: 70 8 (775.0% bigger)
$ hg cat -r ".^" x $ hg cat -r ".^" x
x x
x x
x x
x x

tests/test-remotefilelog-repack.t

$ hg revert x $ hg revert x
$ hg commit -m 'add y back without metadata' $ hg commit -m 'add y back without metadata'
$ cd ../shallow $ cd ../shallow
$ hg pull -q $ hg pull -q
$ hg up -q tip $ hg up -q tip
2 files fetched over 2 fetches - (2 misses, 0.00% hit ratio) over * (glob) 2 files fetched over 2 fetches - (2 misses, 0.00% hit ratio) over * (glob)
$ hg repack $ hg repack
$ ls $TESTTMP/hgcache/master/packs $ ls $TESTTMP/hgcache/master/packs
e634f60d2a9539fc595b1f4db480c64556a396c7.dataidx e8fdf7ae22b772dcc291f905b9c6e5f381d28739.dataidx
e634f60d2a9539fc595b1f4db480c64556a396c7.datapack e8fdf7ae22b772dcc291f905b9c6e5f381d28739.datapack
ebbd7411e00456c0eec8d1150a77e2b3ef490f3f.histidx ebbd7411e00456c0eec8d1150a77e2b3ef490f3f.histidx
ebbd7411e00456c0eec8d1150a77e2b3ef490f3f.histpack ebbd7411e00456c0eec8d1150a77e2b3ef490f3f.histpack
$ hg debughistorypack $TESTTMP/hgcache/master/packs/*.histidx $ hg debughistorypack $TESTTMP/hgcache/master/packs/*.histidx
x x
Node P1 Node P2 Node Link Node Copy From Node P1 Node P2 Node Link Node Copy From
cd410a44d584 577959738234 000000000000 609547eda446 y cd410a44d584 577959738234 000000000000 609547eda446 y
1bb2e6237e03 d4a3ed9310e5 000000000000 0b03bbc9e1e7 1bb2e6237e03 d4a3ed9310e5 000000000000 0b03bbc9e1e7
$ rm -rf $TESTTMP/hgcache/master/packs $ rm -rf $TESTTMP/hgcache/master/packs
$ cp -R $TESTTMP/backuppacks $TESTTMP/hgcache/master/packs $ cp -R $TESTTMP/backuppacks $TESTTMP/hgcache/master/packs
# Test repacking datapack without history # Test repacking datapack without history
$ rm -rf $CACHEDIR/master/packs/*hist* $ rm -rf $CACHEDIR/master/packs/*hist*
$ hg repack $ hg repack
$ hg debugdatapack $TESTTMP/hgcache/master/packs/*.datapack $ hg debugdatapack $TESTTMP/hgcache/master/packs/*.datapack
$TESTTMP/hgcache/master/packs/055c02949317b8507cdb7aaf2e00cc00fd0c5716: $TESTTMP/hgcache/master/packs/a8d86ff8e1a11a77a85f5fea567f56a757583eda:
x: x:
Node Delta Base Delta Length Blob Size Node Delta Base Delta Length Blob Size
1bb2e6237e03 000000000000 8 8 1bb2e6237e03 000000000000 8 8
aee31534993a 000000000000 4 4 d4a3ed9310e5 1bb2e6237e03 12 6
d4a3ed9310e5 000000000000 6 6 aee31534993a d4a3ed9310e5 12 4
Total: 18 18 (0.0% bigger) Total: 32 18 (77.8% bigger)
y: y:
Node Delta Base Delta Length Blob Size Node Delta Base Delta Length Blob Size
577959738234 000000000000 70 8 577959738234 000000000000 70 8
Total: 70 8 (775.0% bigger) Total: 70 8 (775.0% bigger)
$ hg cat -r ".^" x $ hg cat -r ".^" x
x x