Mercurial > hg
view tests/test-dirstate-race.t @ 44363:f7459da77f23
nodemap: introduce an option to use mmap to read the nodemap mapping
The performance and memory benefit is much greater if we don't have to copy all
the data in memory for each information. So we introduce an option (on by
default) to read the data using mmap.
This changeset is the last one definition the API for index support nodemap
data. (they have to be able to use the mmaping).
Below are some benchmark comparing the best we currently have in 5.3 with the
final step of this series (using the persistent nodemap implementation in
Rust). The benchmark run `hg perfindex` with various revset and the following
variants:
Before:
* do not use the persistent nodemap
* use the CPython implementation of the index for nodemap
* use mmapping of the changelog index
After:
* use the MixedIndex Rust code, with the NodeTree object for nodemap access
(still in review)
* use the persistent nodemap data from disk
* access the persistent nodemap data through mmap
* use mmapping of the changelog index
The persistent nodemap greatly speed up most operation on very large
repositories. Some of the previously very fast lookup end up a bit slower because
the persistent nodemap has to be setup. However the absolute slowdown is very
small and won't matters in the big picture.
Here are some numbers (in seconds) for the reference copy of mozilla-try:
Revset Before After abs-change speedup
-10000: 0.004622 0.005532 0.000910 × 0.83
-10: 0.000050 0.000132 0.000082 × 0.37
tip 0.000052 0.000085 0.000033 × 0.61
0 + (-10000:) 0.028222 0.005337 -0.022885 × 5.29
0 0.023521 0.000084 -0.023437 × 280.01
(-10000:) + 0 0.235539 0.005308 -0.230231 × 44.37
(-10:) + :9 0.232883 0.000180 -0.232703 ×1293.79
(-10000:) + (:99) 0.238735 0.005358 -0.233377 × 44.55
:99 + (-10000:) 0.317942 0.005593 -0.312349 × 56.84
:9 + (-10:) 0.313372 0.000179 -0.313193 ×1750.68
:9 0.316450 0.000143 -0.316307 ×2212.93
On smaller repositories, the cost of nodemap related operation is not as big, so
the win is much more modest. Yet it helps shaving a handful of millisecond here
and there.
Here are some numbers (in seconds) for the reference copy of mercurial:
Revset Before After abs-change speedup
-10: 0.000065 0.000097 0.000032 × 0.67
tip 0.000063 0.000078 0.000015 × 0.80
0 0.000561 0.000079 -0.000482 × 7.10
-10000: 0.004609 0.003648 -0.000961 × 1.26
0 + (-10000:) 0.005023 0.003715 -0.001307 × 1.35
(-10:) + :9 0.002187 0.000108 -0.002079 ×20.25
(-10000:) + 0 0.006252 0.003716 -0.002536 × 1.68
(-10000:) + (:99) 0.006367 0.003707 -0.002660 × 1.71
:9 + (-10:) 0.003846 0.000110 -0.003736 ×34.96
:9 0.003854 0.000099 -0.003755 ×38.92
:99 + (-10000:) 0.007644 0.003778 -0.003866 × 2.02
Differential Revision: https://phab.mercurial-scm.org/D7894
author | Pierre-Yves David <pierre-yves.david@octobus.net> |
---|---|
date | Tue, 11 Feb 2020 11:18:52 +0100 |
parents | 28a4fb793ba1 |
children | f90a5c211251 |
line wrap: on
line source
$ 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 co -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 >> $TESTTMP/dirstaterace.py << EOF > from mercurial import ( > context, > extensions, > ) > def extsetup(ui): > extensions.wrapfunction(context.workingctx, '_checklookup', overridechecklookup) > def overridechecklookup(orig, self, files): > # make an update that changes the dirstate from underneath > self._repo.ui.system(br"sh '$TESTTMP/dirstaterace.sh'", > cwd=self._repo.root) > return orig(self, files) > 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 $ 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 > 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 "c4" (tip) 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