pathcopies: give up any optimization based on `introrev`
Between
8a0136f69027 and
d98fb3f42f33, we sped up the search for the
introduction revision during path copies. However, further checking show that
finding the introduction revision is still expensive and that we are better off
without it. So we simply drop it and only rely on the linkrev optimisation.
I ran `perfpathcopies` on 6989 pair of revision in the pypy
repository (`hg perfhelper-pathcopies`. The result is massively in favor of
dropping this condition. The result of the copy tracing are unchanged.
Attempt to use a smaller changes preserving linkrev usage were unsuccessful, it
can return wrong result. The following changesets broke test-mv-cp-st-diff.t
- if not f.isintroducedafter(limit):
+ if limit >= 0 and f.linkrev() < limit:
return None
Here are various numbers (before this changeset/after this changesets)
source destination before after saved-time ratio
worth cases
e66f24650daf 695dfb0f493b 1.062843 1.246369 -0.183526 1.172675
c979853a3b6a 8d60fe293e79 1.036985 1.196414 -0.159429 1.153743
22349fa2fc33 fbb1c9fd86c0 0.879926 1.038682 -0.158756 1.180420
682b98f3e672 a4878080a536 0.909952 1.063801 -0.153849 1.169074
5adabc9b9848 920958a93997 0.993622 1.147452 -0.153830 1.154817
worse 1%
dbfbfcf077e9 aea8f2fd3593 1.016595 1.082999 -0.066404 1.065320
worse 5%
c95f1ced15f2 7d29d5e39734 0.453694 0.471156 -0.017462 1.038488
worse 10%
3e144ed1d5b7 2aef0e942480 0.035140 0.037535 -0.002395 1.068156
worse 25%
321fc60db035 801748ba582a 0.009267 0.009325 -0.000058 1.006259
median
2088ce763fc2 e6991321d78b 0.000665 0.000651 0.000014 0.978947
best 25%
915631a97de6 385b31354be6 0.040743 0.040363 0.000380 0.990673
best 10%
ad495c36a765 19c10384d3e7 0.431658 0.411490 0.020168 0.953278
best 5%
d13ae7d283ae 813c99f810ac 1.141404 1.075346 0.066058 0.942126
best 1%
81593cb4a496 99ae11866969 1.833297 0.063823 1.769474 0.034813
best cases
c3b14617fbd7 743a0fcaa4eb 1101.811740 2.735970 1099.075770 0.002483
c3b14617fbd7 9ba6ab77fd29 1116.753953 2.800729 1113.953224 0.002508
058b99d6e81f 57e249b7a3ea 1246.128485 3.042762 1243.085723 0.002442
9a8c361aab49 0354a250d371 1253.111894 3.085796 1250.026098 0.002463
442dbbc53c68 3ec1002a818c 1261.786294 3.138607 1258.647687 0.002487
As one can see, the average case is not really impacted. However, the worth case
we get after this changeset are much better than the one we had before it. We
have 30 pairs where improvements are above 10 minutes.
This reflect in the combined time for all pairs
before: 26256s
after: 1300s (-95%)
If we remove these pathological 30 cases, we still see a significant improvements:
before: 1631s
after: 1245s (-24%)
#require hardlink reporevlogstore
$ cat > nlinks.py <<EOF
> from __future__ import print_function
> import sys
> from mercurial import pycompat, util
> for f in sorted(sys.stdin.readlines()):
> f = f[:-1]
> print(util.nlinks(pycompat.fsencode(f)), f)
> EOF
$ nlinksdir()
> {
> find "$@" -type f | "$PYTHON" $TESTTMP/nlinks.py
> }
Some implementations of cp can't create hardlinks (replaces 'cp -al' on Linux):
$ cat > linkcp.py <<EOF
> from __future__ import absolute_import
> import sys
> from mercurial import pycompat, util
> util.copyfiles(pycompat.fsencode(sys.argv[1]),
> pycompat.fsencode(sys.argv[2]), hardlink=True)
> EOF
$ linkcp()
> {
> "$PYTHON" $TESTTMP/linkcp.py $1 $2
> }
Prepare repo r1:
$ hg init r1
$ cd r1
$ echo c1 > f1
$ hg add f1
$ hg ci -m0
$ mkdir d1
$ cd d1
$ echo c2 > f2
$ hg add f2
$ hg ci -m1
$ cd ../..
$ nlinksdir r1/.hg/store
1 r1/.hg/store/00changelog.i
1 r1/.hg/store/00manifest.i
1 r1/.hg/store/data/d1/f2.i
1 r1/.hg/store/data/f1.i
1 r1/.hg/store/fncache (repofncache !)
1 r1/.hg/store/phaseroots
1 r1/.hg/store/undo
1 r1/.hg/store/undo.backup.fncache (repofncache !)
1 r1/.hg/store/undo.backupfiles
1 r1/.hg/store/undo.phaseroots
Create hardlinked clone r2:
$ hg clone -U --debug r1 r2 --config progress.debug=true
linking: 1 files
linking: 2 files
linking: 3 files
linking: 4 files
linking: 5 files
linking: 6 files
linking: 7 files
linked 7 files
Create non-hardlinked clone r3:
$ hg clone --pull r1 r3
requesting all changes
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
new changesets 40d85e9847f2:7069c422939c
updating to branch default
2 files updated, 0 files merged, 0 files removed, 0 files unresolved
Repos r1 and r2 should now contain hardlinked files:
$ nlinksdir r1/.hg/store
2 r1/.hg/store/00changelog.i
2 r1/.hg/store/00manifest.i
2 r1/.hg/store/data/d1/f2.i
2 r1/.hg/store/data/f1.i
2 r1/.hg/store/fncache (repofncache !)
1 r1/.hg/store/phaseroots
1 r1/.hg/store/undo
1 r1/.hg/store/undo.backup.fncache (repofncache !)
1 r1/.hg/store/undo.backupfiles
1 r1/.hg/store/undo.phaseroots
$ nlinksdir r2/.hg/store
2 r2/.hg/store/00changelog.i
2 r2/.hg/store/00manifest.i
2 r2/.hg/store/data/d1/f2.i
2 r2/.hg/store/data/f1.i
2 r2/.hg/store/fncache (repofncache !)
Repo r3 should not be hardlinked:
$ nlinksdir r3/.hg/store
1 r3/.hg/store/00changelog.i
1 r3/.hg/store/00manifest.i
1 r3/.hg/store/data/d1/f2.i
1 r3/.hg/store/data/f1.i
1 r3/.hg/store/fncache (repofncache !)
1 r3/.hg/store/phaseroots
1 r3/.hg/store/undo
1 r3/.hg/store/undo.backupfiles
1 r3/.hg/store/undo.phaseroots
Create a non-inlined filelog in r3:
$ cd r3/d1
>>> f = open('data1', 'wb')
>>> for x in range(10000):
... f.write(b"%d\n" % x) and None
>>> f.close()
$ for j in 0 1 2 3 4 5 6 7 8 9; do
> cat data1 >> f2
> hg commit -m$j
> done
$ cd ../..
$ nlinksdir r3/.hg/store
1 r3/.hg/store/00changelog.i
1 r3/.hg/store/00manifest.i
1 r3/.hg/store/data/d1/f2.d
1 r3/.hg/store/data/d1/f2.i
1 r3/.hg/store/data/f1.i
1 r3/.hg/store/fncache (repofncache !)
1 r3/.hg/store/phaseroots
1 r3/.hg/store/undo
1 r3/.hg/store/undo.backup.fncache (repofncache !)
1 r3/.hg/store/undo.backup.phaseroots
1 r3/.hg/store/undo.backupfiles
1 r3/.hg/store/undo.phaseroots
Push to repo r1 should break up most hardlinks in r2:
$ hg -R r2 verify
checking changesets
checking manifests
crosschecking files in changesets and manifests
checking files
checked 2 changesets with 2 changes to 2 files
$ cd r3
$ hg push
pushing to $TESTTMP/r1
searching for changes
adding changesets
adding manifests
adding file changes
added 10 changesets with 10 changes to 1 files
$ cd ..
$ nlinksdir r2/.hg/store
1 r2/.hg/store/00changelog.i
1 r2/.hg/store/00manifest.i
1 r2/.hg/store/data/d1/f2.i
2 r2/.hg/store/data/f1.i
[12] r2/\.hg/store/fncache (re) (repofncache !)
#if hardlink-whitelisted repofncache
$ nlinksdir r2/.hg/store/fncache
2 r2/.hg/store/fncache
#endif
$ hg -R r2 verify
checking changesets
checking manifests
crosschecking files in changesets and manifests
checking files
checked 2 changesets with 2 changes to 2 files
$ cd r1
$ hg up
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
Committing a change to f1 in r1 must break up hardlink f1.i in r2:
$ echo c1c1 >> f1
$ hg ci -m00
$ cd ..
$ nlinksdir r2/.hg/store
1 r2/.hg/store/00changelog.i
1 r2/.hg/store/00manifest.i
1 r2/.hg/store/data/d1/f2.i
1 r2/.hg/store/data/f1.i
[12] r2/\.hg/store/fncache (re) (repofncache !)
#if hardlink-whitelisted repofncache
$ nlinksdir r2/.hg/store/fncache
2 r2/.hg/store/fncache
#endif
Create a file which exec permissions we will change
$ cd r3
$ echo "echo hello world" > f3
$ hg add f3
$ hg ci -mf3
$ cd ..
$ cd r3
$ hg tip --template '{rev}:{node|short}\n'
12:d3b77733a28a
$ echo bla > f1
$ chmod +x f3
$ hg ci -m1
$ cd ..
Create hardlinked copy r4 of r3 (on Linux, we would call 'cp -al'):
$ linkcp r3 r4
'checklink' is produced by hardlinking a symlink, which is undefined whether
the symlink should be followed or not. It does behave differently on Linux and
BSD. Just remove it so the test pass on both platforms.
$ rm -f r4/.hg/wcache/checklink
r4 has hardlinks in the working dir (not just inside .hg):
$ nlinksdir r4
2 r4/.hg/00changelog.i
2 r4/.hg/branch
2 r4/.hg/cache/branch2-base
2 r4/.hg/cache/branch2-served
2 r4/.hg/cache/rbc-names-v1
2 r4/.hg/cache/rbc-revs-v1
2 r4/.hg/dirstate
2 r4/.hg/fsmonitor.state (fsmonitor !)
2 r4/.hg/hgrc
2 r4/.hg/last-message.txt
2 r4/.hg/requires
2 r4/.hg/store/00changelog.i
2 r4/.hg/store/00manifest.i
2 r4/.hg/store/data/d1/f2.d
2 r4/.hg/store/data/d1/f2.i
2 r4/.hg/store/data/f1.i
2 r4/.hg/store/data/f3.i
2 r4/.hg/store/fncache (repofncache !)
2 r4/.hg/store/phaseroots
2 r4/.hg/store/undo
2 r4/.hg/store/undo.backup.fncache (repofncache !)
2 r4/.hg/store/undo.backup.phaseroots
2 r4/.hg/store/undo.backupfiles
2 r4/.hg/store/undo.phaseroots
[24] r4/\.hg/undo\.backup\.dirstate (re)
2 r4/.hg/undo.bookmarks
2 r4/.hg/undo.branch
2 r4/.hg/undo.desc
[24] r4/\.hg/undo\.dirstate (re)
2 r4/.hg/wcache/checkisexec (execbit !)
2 r4/.hg/wcache/checklink-target (symlink !)
2 r4/.hg/wcache/checknoexec (execbit !)
2 r4/.hg/wcache/manifestfulltextcache (reporevlogstore !)
2 r4/d1/data1
2 r4/d1/f2
2 r4/f1
2 r4/f3
Update back to revision 12 in r4 should break hardlink of file f1 and f3:
#if hardlink-whitelisted
$ nlinksdir r4/.hg/undo.backup.dirstate r4/.hg/undo.dirstate
4 r4/.hg/undo.backup.dirstate
4 r4/.hg/undo.dirstate
#endif
$ hg -R r4 up 12
2 files updated, 0 files merged, 0 files removed, 0 files unresolved (execbit !)
1 files updated, 0 files merged, 0 files removed, 0 files unresolved (no-execbit !)
$ nlinksdir r4
2 r4/.hg/00changelog.i
1 r4/.hg/branch
2 r4/.hg/cache/branch2-base
2 r4/.hg/cache/branch2-served
2 r4/.hg/cache/rbc-names-v1
2 r4/.hg/cache/rbc-revs-v1
1 r4/.hg/dirstate
1 r4/.hg/fsmonitor.state (fsmonitor !)
2 r4/.hg/hgrc
2 r4/.hg/last-message.txt
2 r4/.hg/requires
2 r4/.hg/store/00changelog.i
2 r4/.hg/store/00manifest.i
2 r4/.hg/store/data/d1/f2.d
2 r4/.hg/store/data/d1/f2.i
2 r4/.hg/store/data/f1.i
2 r4/.hg/store/data/f3.i
2 r4/.hg/store/fncache
2 r4/.hg/store/phaseroots
2 r4/.hg/store/undo
2 r4/.hg/store/undo.backup.fncache (repofncache !)
2 r4/.hg/store/undo.backup.phaseroots
2 r4/.hg/store/undo.backupfiles
2 r4/.hg/store/undo.phaseroots
[24] r4/\.hg/undo\.backup\.dirstate (re)
2 r4/.hg/undo.bookmarks
2 r4/.hg/undo.branch
2 r4/.hg/undo.desc
[24] r4/\.hg/undo\.dirstate (re)
2 r4/.hg/wcache/checkisexec (execbit !)
2 r4/.hg/wcache/checklink-target (symlink !)
2 r4/.hg/wcache/checknoexec (execbit !)
1 r4/.hg/wcache/manifestfulltextcache (reporevlogstore !)
2 r4/d1/data1
2 r4/d1/f2
1 r4/f1
1 r4/f3 (execbit !)
2 r4/f3 (no-execbit !)
#if hardlink-whitelisted
$ nlinksdir r4/.hg/undo.backup.dirstate r4/.hg/undo.dirstate
4 r4/.hg/undo.backup.dirstate
4 r4/.hg/undo.dirstate
#endif
Test hardlinking outside hg:
$ mkdir x
$ echo foo > x/a
$ linkcp x y
$ echo bar >> y/a
No diff if hardlink:
$ diff x/a y/a
Test mq hardlinking:
$ echo "[extensions]" >> $HGRCPATH
$ echo "mq=" >> $HGRCPATH
$ hg init a
$ cd a
$ hg qimport -n foo - << EOF
> # HG changeset patch
> # Date 1 0
> diff -r 2588a8b53d66 a
> --- /dev/null Thu Jan 01 00:00:00 1970 +0000
> +++ b/a Wed Jul 23 15:54:29 2008 +0200
> @@ -0,0 +1,1 @@
> +a
> EOF
adding foo to series file
$ hg qpush
applying foo
now at: foo
$ cd ..
$ linkcp a b
$ cd b
$ hg qimport -n bar - << EOF
> # HG changeset patch
> # Date 2 0
> diff -r 2588a8b53d66 a
> --- /dev/null Thu Jan 01 00:00:00 1970 +0000
> +++ b/b Wed Jul 23 15:54:29 2008 +0200
> @@ -0,0 +1,1 @@
> +b
> EOF
adding bar to series file
$ hg qpush
applying bar
now at: bar
$ cat .hg/patches/status
430ed4828a74fa4047bc816a25500f7472ab4bfe:foo
4e7abb4840c46a910f6d7b4d3c3fc7e5209e684c:bar
$ cat .hg/patches/series
foo
bar
$ cat ../a/.hg/patches/status
430ed4828a74fa4047bc816a25500f7472ab4bfe:foo
$ cat ../a/.hg/patches/series
foo
Test tags hardlinking:
$ hg qdel -r qbase:qtip
patch foo finalized without changeset message
patch bar finalized without changeset message
$ hg tag -l lfoo
$ hg tag foo
$ cd ..
$ linkcp b c
$ cd c
$ hg tag -l -r 0 lbar
$ hg tag -r 0 bar
$ cat .hgtags
4e7abb4840c46a910f6d7b4d3c3fc7e5209e684c foo
430ed4828a74fa4047bc816a25500f7472ab4bfe bar
$ cat .hg/localtags
4e7abb4840c46a910f6d7b4d3c3fc7e5209e684c lfoo
430ed4828a74fa4047bc816a25500f7472ab4bfe lbar
$ cat ../b/.hgtags
4e7abb4840c46a910f6d7b4d3c3fc7e5209e684c foo
$ cat ../b/.hg/localtags
4e7abb4840c46a910f6d7b4d3c3fc7e5209e684c lfoo
$ cd ..