annotate: pre-calculate the "needed" dictionary (issue5360)
The "needed" dict is used as a reference counter to free items in the giant
"hist" dict. However, currently it is not very accurate and can lead to
dropping "hist" items unnecessarily, for example, with the following DAG,
-3-
/ \
0--1--2--4--
The current algorithm will visit and calculate rev 1 twice, undesired. And
it tries to be smart by clearing rev 1's parents: "pcache[1] = []" at the
time hist[1] being accessed (note: hist[1] needs to be used twice, by rev 2
and rev 3). It can result in incorrect results if p1 of rev 4 deletes chunks
belonging to rev 0.
However, simply removing "needed" is not okay, because it will consume 10x
memory:
# without any change
% HGRCPATH= lrun ./hg annotate mercurial/commands.py -r d130a38 3>&2 [1]
MEMORY 49074176
CPUTIME 9.213
REALTIME 9.270
# with "needed" removed
MEMORY 637673472
CPUTIME 8.164
REALTIME 8.249
This patch moves "needed" (and "pcache") calculation to a separate DFS to
address the issue. It improves perf and fixes issue5360 by correctly reusing
hist, while maintaining low memory usage. Some additional attempt has been
made to further reduce memory usage, like changing "pcache[f] = []" to "del
pcache[f]". Therefore the result can be both faster and lower memory usage:
# with this patch applied
MEMORY 47575040
CPUTIME 7.870
REALTIME 7.926
[1]: lrun is a lightweight sandbox built on Linux cgroup and namespace. It's
used to measure CPU and memory usage here. Source code is available at
github.com/quark-zju/lrun.
#require cvs
$ filterpath()
> {
> eval "$@" | sed "s:$CVSROOT:*REPO*:g"
> }
$ cvscall()
> {
> cvs -f "$@"
> }
output of 'cvs ci' varies unpredictably, so discard most of it
-- just keep the part that matters
$ cvsci()
> {
> cvs -f ci -f "$@" > /dev/null
> }
$ hgcat()
> {
> hg --cwd src-hg cat -r tip "$1"
> }
$ echo "[extensions]" >> $HGRCPATH
$ echo "convert = " >> $HGRCPATH
create cvs repository
$ mkdir cvsmaster
$ cd cvsmaster
$ CVSROOT=`pwd`
$ export CVSROOT
$ CVS_OPTIONS=-f
$ export CVS_OPTIONS
$ cd ..
$ rmdir cvsmaster
$ filterpath cvscall -Q -d "$CVSROOT" init
checkout #1: add foo.txt
$ cvscall -Q checkout -d cvsworktmp .
$ cd cvsworktmp
$ mkdir foo
$ cvscall -Q add foo
$ cd foo
$ echo foo > foo.txt
$ cvscall -Q add foo.txt
$ cvsci -m "add foo.txt" foo.txt
$ cd ../..
$ rm -rf cvsworktmp
checkout #2: create MYBRANCH1 and modify foo.txt on it
$ cvscall -Q checkout -d cvswork foo
$ cd cvswork
$ cvscall -q rtag -b -R MYBRANCH1 foo
$ cvscall -Q update -P -r MYBRANCH1
$ echo bar > foo.txt
$ cvsci -m "bar" foo.txt
$ echo baz > foo.txt
$ cvsci -m "baz" foo.txt
create MYBRANCH1_2 and modify foo.txt some more
$ cvscall -q rtag -b -R -r MYBRANCH1 MYBRANCH1_2 foo
$ cvscall -Q update -P -r MYBRANCH1_2
$ echo bazzie > foo.txt
$ cvsci -m "bazzie" foo.txt
create MYBRANCH1_1 and modify foo.txt yet again
$ cvscall -q rtag -b -R MYBRANCH1_1 foo
$ cvscall -Q update -P -r MYBRANCH1_1
$ echo quux > foo.txt
$ cvsci -m "quux" foo.txt
merge MYBRANCH1 to MYBRANCH1_1
$ filterpath cvscall -Q update -P -jMYBRANCH1
rcsmerge: warning: conflicts during merge
RCS file: *REPO*/foo/foo.txt,v
retrieving revision 1.1
retrieving revision 1.1.2.2
Merging differences between 1.1 and 1.1.2.2 into foo.txt
carefully placed sleep to dodge cvs bug (optimization?) where it
sometimes ignores a "commit" command if it comes too fast (the -f
option in cvsci seems to work for all the other commits in this
script)
$ sleep 1
$ echo xyzzy > foo.txt
$ cvsci -m "merge1+clobber" foo.txt
#if unix-permissions
return to trunk and merge MYBRANCH1_2
$ cvscall -Q update -P -A
$ filterpath cvscall -Q update -P -jMYBRANCH1_2
RCS file: *REPO*/foo/foo.txt,v
retrieving revision 1.1
retrieving revision 1.1.2.2.2.1
Merging differences between 1.1 and 1.1.2.2.2.1 into foo.txt
$ cvsci -m "merge2" foo.txt
$ REALCVS=`which cvs`
$ echo "for x in \$*; do if [ \"\$x\" = \"rlog\" ]; then echo \"RCS file: $CVSROOT/foo/foo.txt,v\"; cat \"$TESTDIR/test-convert-cvsnt-mergepoints.rlog\"; exit 0; fi; done; $REALCVS \$*" > ../cvs
$ chmod +x ../cvs
$ PATH=..:${PATH} hg debugcvsps --parents foo
collecting CVS rlog
7 log entries
creating changesets
7 changeset entries
---------------------
PatchSet 1
Date: * (glob)
Author: user
Branch: HEAD
Tag: (none)
Branchpoints: MYBRANCH1, MYBRANCH1_1
Log:
foo.txt
Members:
foo.txt:INITIAL->1.1
---------------------
PatchSet 2
Date: * (glob)
Author: user
Branch: MYBRANCH1
Tag: (none)
Parent: 1
Log:
bar
Members:
foo.txt:1.1->1.1.2.1
---------------------
PatchSet 3
Date: * (glob)
Author: user
Branch: MYBRANCH1
Tag: (none)
Branchpoints: MYBRANCH1_2
Parent: 2
Log:
baz
Members:
foo.txt:1.1.2.1->1.1.2.2
---------------------
PatchSet 4
Date: * (glob)
Author: user
Branch: MYBRANCH1_1
Tag: (none)
Parent: 1
Log:
quux
Members:
foo.txt:1.1->1.1.4.1
---------------------
PatchSet 5
Date: * (glob)
Author: user
Branch: MYBRANCH1_2
Tag: (none)
Parent: 3
Log:
bazzie
Members:
foo.txt:1.1.2.2->1.1.2.2.2.1
---------------------
PatchSet 6
Date: * (glob)
Author: user
Branch: HEAD
Tag: (none)
Parents: 1,5
Log:
merge
Members:
foo.txt:1.1->1.2
---------------------
PatchSet 7
Date: * (glob)
Author: user
Branch: MYBRANCH1_1
Tag: (none)
Parents: 4,3
Log:
merge
Members:
foo.txt:1.1.4.1->1.1.4.2
#endif
$ cd ..