Mercurial > hg
view tests/test-dirstate-race.t @ 50671:e06d1a779eb6
store: stop relying on a `revlog_type` property
We want to know if a file is related to a revlog, but the rest is dealt with
differently already, so we simplify things further.
as a bonus, this cleanup This provides a small but noticeable speedup.
The number below use `hg perf::stream-locked-section` to measure the time spend
in the locked section of the streaming clone. Number are run on various
repository and compare different steps.:
1) the effect of this patchs,
2) the effect of the cleanup series,
2) current state compared to because large refactoring.
All benchmarks are run on linux with Python 3.10.7.
### Effect of this patch
# mercurial-2018-08-01-zstd-sparse-revlog
# benchmark.name = perf-stream-locked-section
before: 0.030246 seconds
after: 0.029274 seconds (-3.21%)
# pypy-2018-08-01-zstd-sparse-revlog
before: 0.545012 seconds
after: 0.520872 seconds (-4.43%)
# netbeans-2018-08-01-zstd-sparse-revlog
before: 2.719939 seconds
after: 2.626791 seconds (-3.42%)
# mozilla-central-2018-08-01-zstd-sparse-revlog
before: 6.304179 seconds
after: 6.096700 seconds (-3.29%)
# mozilla-try-2019-02-18-zstd-sparse-revlog
before: 14.142687 seconds
after: 13.640779 seconds (-3.55%)
### Effect of this series
A small but sizeable speedup
# mercurial-2018-08-01-zstd-sparse-revlog
before: 0.031122 seconds
after: 0.029274 seconds (-5.94%)
# pypy-2018-08-01-zstd-sparse-revlog
before: 0.589970 seconds
after: 0.520872 seconds (-11.71%)
# netbeans-2018-08-01-zstd-sparse-revlog
before: 2.980300 seconds
after: 2.626791 seconds (-11.86%)
# mozilla-central-2018-08-01-zstd-sparse-revlog
before: 6.863204 seconds
after: 6.096700 seconds (-11.17%)
# mozilla-try-2019-02-18-zstd-sparse-revlog
before: 14.921393 seconds
after: 13.640779 seconds (-8.58%)
### Current state compared to the pre-refactoring state
The refactoring introduced multiple string manipulation and dictionary creation
that seems to induce a signifiant slowdown
Slowdown
# mercurial-2018-08-01-zstd-sparse-revlog
6.4.3: 0.019459 seconds
after: 0.029274 seconds (+50.44%)
## pypy-2018-08-01-zstd-sparse-revlog
6.4.3: 0.290715 seconds
after: 0.520872 seconds (+79.17%)
# netbeans-2018-08-01-zstd-sparse-revlog
6.4.3: 1.403447 seconds
after: 2.626791 seconds (+87.17%)
# mozilla-central-2018-08-01-zstd-sparse-revlog
6.4.3: 3.163549 seconds
after: 6.096700 seconds (+92.72%)
# mozilla-try-2019-02-18-zstd-sparse-revlog
6.4.3: 6.702184 seconds
after: 13.640779 seconds (+103.53%)
author | Pierre-Yves David <pierre-yves.david@octobus.net> |
---|---|
date | Tue, 30 May 2023 17:43:59 +0100 |
parents | e845537f6adb |
children |
line wrap: on
line source
#testcases dirstate-v1 dirstate-v2 #if dirstate-v2 $ cat >> $HGRCPATH << EOF > [format] > use-dirstate-v2=1 > [storage] > dirstate-v2.slow-path=allow > EOF #endif $ hg init repo $ cd repo $ echo a > a $ hg add a $ hg commit -m test Do we ever miss a sub-second change?: $ for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20; do > hg update -qC 0 > echo b > a > hg st > done M a M a M a M a M a M a M a M a M a M a M a M a M a M a M a M a M a M a M a M a $ echo test > b $ mkdir dir1 $ echo test > dir1/c $ echo test > d $ echo test > e #if execbit A directory will typically have the execute bit -- make sure it doesn't get confused with a file with the exec bit set $ chmod +x e #endif $ hg add b dir1 d e adding dir1/c $ hg commit -m test2 $ cat >> $TESTTMP/dirstaterace.py << EOF > from mercurial import ( > context, > extensions, > ) > def extsetup(ui): > extensions.wrapfunction(context.workingctx, '_checklookup', overridechecklookup) > def overridechecklookup(orig, self, *args, **kwargs): > # make an update that changes the dirstate from underneath > self._repo.ui.system(br"sh '$TESTTMP/dirstaterace.sh'", > cwd=self._repo.root) > return orig(self, *args, **kwargs) > EOF $ hg debugrebuilddirstate $ hg debugdirstate n 0 -1 unset a n 0 -1 unset b n 0 -1 unset d n 0 -1 unset dir1/c n 0 -1 unset e XXX Note that this returns M for files that got replaced by directories. This is definitely a bug, but the fix for that is hard and the next status run is fine anyway. $ cat > $TESTTMP/dirstaterace.sh <<EOF > rm b && rm -r dir1 && rm d && mkdir d && rm e && mkdir e > EOF $ sleep 1 # ensure non-ambiguous mtime $ hg status --config extensions.dirstaterace=$TESTTMP/dirstaterace.py M d M e ! b ! dir1/c $ hg debugdirstate n 644 2 * a (glob) n 0 -1 unset b n 0 -1 unset d n 0 -1 unset dir1/c n 0 -1 unset e $ hg status ! b ! d ! dir1/c ! e $ rmdir d e $ hg update -C -q . Test that dirstate changes aren't written out at the end of "hg status", if .hg/dirstate is already changed simultaneously before acquisition of wlock in workingctx._poststatusfixup(). This avoidance is important to keep consistency of dirstate in race condition (see issue5584 for detail). $ hg parents -q 1:* (glob) $ hg debugrebuilddirstate $ hg debugdirstate n 0 -1 unset a n 0 -1 unset b n 0 -1 unset d n 0 -1 unset dir1/c n 0 -1 unset e $ cat > $TESTTMP/dirstaterace.sh <<EOF > # This script assumes timetable of typical issue5584 case below: > # > # 1. "hg status" loads .hg/dirstate > # 2. "hg status" confirms clean-ness of FILE > # 3. "hg update -C 0" updates the working directory simultaneously > # (FILE is removed, and FILE is dropped from .hg/dirstate) > # 4. "hg status" acquires wlock > # (.hg/dirstate is re-loaded = no FILE entry in dirstate) > # 5. "hg status" marks FILE in dirstate as clean > # (FILE entry is added to in-memory dirstate) > # 6. "hg status" writes dirstate changes into .hg/dirstate > # (FILE entry is written into .hg/dirstate) > # > # To reproduce similar situation easily and certainly, #2 and #3 > # are swapped. "hg cat" below ensures #2 on "hg status" side. > > hg update -q -C 0 > hg cat -r 1 b > b > # make sure the timestamps is not ambiguous and a write will be issued > touch -t 198606251012 b > EOF "hg status" below should excludes "e", of which exec flag is set, for portability of test scenario, because unsure but missing "e" is treated differently in _checklookup() according to runtime platform. - "missing(!)" on POSIX, "pctx[f].cmp(self[f])" raises ENOENT - "modified(M)" on Windows, "self.flags(f) != pctx.flags(f)" is True $ hg status --config extensions.dirstaterace=$TESTTMP/dirstaterace.py --debug -X path:e skip updating dirstate: identity mismatch M a ! d ! dir1/c $ hg parents -q 0:* (glob) $ hg files a $ hg debugdirstate n * * * a (glob) $ rm b #if fsmonitor Create fsmonitor state. $ hg status $ f --type .hg/fsmonitor.state .hg/fsmonitor.state: file Test that invalidating fsmonitor state in the middle (which doesn't require the wlock) causes the fsmonitor update to be skipped. hg debugrebuilddirstate ensures that the dirstaterace hook will be called, but it also invalidates the fsmonitor state. So back it up and restore it. $ mv .hg/fsmonitor.state .hg/fsmonitor.state.tmp $ hg debugrebuilddirstate $ mv .hg/fsmonitor.state.tmp .hg/fsmonitor.state $ cat > $TESTTMP/dirstaterace.sh <<EOF > rm .hg/fsmonitor.state > EOF $ hg status --config extensions.dirstaterace=$TESTTMP/dirstaterace.py --debug skip updating fsmonitor.state: identity mismatch $ f .hg/fsmonitor.state .hg/fsmonitor.state: file not found #endif Set up a rebase situation for issue5581. $ echo c2 > a $ echo c2 > b $ hg add b $ hg commit -m c2 created new head $ echo c3 >> a $ hg commit -m c3 $ hg update 2 1 files updated, 0 files merged, 0 files removed, 0 files unresolved $ echo c4 >> a $ echo c4 >> b $ hg commit -m c4 created new head Configure a merge tool that runs status in the middle of the rebase. The goal of the status call is to trigger a potential bug if fsmonitor's state is written even though the wlock is held by another process. The output of 'hg status' in the merge tool goes to /dev/null because we're more interested in the results of 'hg status' run after the rebase. $ cat >> $TESTTMP/mergetool-race.sh << EOF > echo "custom merge tool" > printf "c2\nc3\nc4\n" > \$1 > hg --cwd "$TESTTMP/repo" status > /dev/null > echo "custom merge tool end" > EOF $ cat >> $HGRCPATH << EOF > [extensions] > rebase = > [merge-tools] > test.executable=sh > test.args=$TESTTMP/mergetool-race.sh \$output > EOF $ hg rebase -s . -d 3 --tool test rebasing 4:b08445fd6b2a tip "c4" merging a custom merge tool custom merge tool end saved backup bundle to $TESTTMP/repo/.hg/strip-backup/* (glob) This hg status should be empty, whether or not fsmonitor is enabled (issue5581). $ hg status