revlog: don't flush data file after every added revision
The current behavior of revlogs is to flush the data file when writing
data to it. Tracing system calls revealed that changegroup processing
incurred numerous write(2) calls for values much smaller than the
default buffer size (Python defaults to 4096, but it can be adjusted
based on detected block size at run time by CPython).
The reason we flush revlogs is so readers have all data available.
For example, the current code in revlog.py will re-open the revlog
file (instead of seeking an existing file handle) to read the text
of a revision. This happens when starting a new delta chain when
adding several revisions from changegroups, for example. Yes, this
is likely sub-optimal (we should probably be sharing file descriptors
between readers and writers to avoid the flushing and associated
overhead of re-opening files).
While flushing revlogs is necessary, it appears all callers are
diligent about flushing files before a read is performed (see
buildtext() in _addrevision()), making the flush in
_writeentry() redundant and unncessary. So, we remove it. In practice,
this means we incur a write(2) a) when the buffer is full (typically
4096 bytes) b) when a new delta chain is created rather than after
every added revision. This applies to every revlog, but by volume
it mostly impacts filelogs.
Removing the redundant flush from _writeentry() significantly
reduces the number of write(2) calls during changegroup processing on
my Linux machine. When applying a changegroup of the hg repo based on
my local repo, the total number of write(2) calls during application
of the mercurial/localrepo.py revlogs dropped from 1,320 to 217 with
this patch applied. Total I/O related system calls dropped from 1,577
to 474.
When unbundling a mozilla-central gzipped bundle (264,403 changesets
with 1,492,215 changes to 222,507 files), total write(2) calls
dropped from 1,252,881 to 827,106 and total system calls dropped from
3,601,259 to 3,178,636 - a reduction of 425,775!
While the system call reduction is significant, it appears
to have no impact on wall time on my Linux and Windows machines. Still,
fewer syscalls is fewer syscalls. Surely this can't hurt. If nothing
else, it makes examining remaining system call usage simpler and opens
the door to experimenting with the performance impact of different
buffer sizes.
$ hg init repo1
$ cd repo1
$ mkdir a b a/1 b/1 b/2
$ touch in_root a/in_a b/in_b a/1/in_a_1 b/1/in_b_1 b/2/in_b_2
hg status in repo root:
$ hg status
? a/1/in_a_1
? a/in_a
? b/1/in_b_1
? b/2/in_b_2
? b/in_b
? in_root
hg status . in repo root:
$ hg status .
? a/1/in_a_1
? a/in_a
? b/1/in_b_1
? b/2/in_b_2
? b/in_b
? in_root
$ hg status --cwd a
? a/1/in_a_1
? a/in_a
? b/1/in_b_1
? b/2/in_b_2
? b/in_b
? in_root
$ hg status --cwd a .
? 1/in_a_1
? in_a
$ hg status --cwd a ..
? 1/in_a_1
? in_a
? ../b/1/in_b_1
? ../b/2/in_b_2
? ../b/in_b
? ../in_root
$ hg status --cwd b
? a/1/in_a_1
? a/in_a
? b/1/in_b_1
? b/2/in_b_2
? b/in_b
? in_root
$ hg status --cwd b .
? 1/in_b_1
? 2/in_b_2
? in_b
$ hg status --cwd b ..
? ../a/1/in_a_1
? ../a/in_a
? 1/in_b_1
? 2/in_b_2
? in_b
? ../in_root
$ hg status --cwd a/1
? a/1/in_a_1
? a/in_a
? b/1/in_b_1
? b/2/in_b_2
? b/in_b
? in_root
$ hg status --cwd a/1 .
? in_a_1
$ hg status --cwd a/1 ..
? in_a_1
? ../in_a
$ hg status --cwd b/1
? a/1/in_a_1
? a/in_a
? b/1/in_b_1
? b/2/in_b_2
? b/in_b
? in_root
$ hg status --cwd b/1 .
? in_b_1
$ hg status --cwd b/1 ..
? in_b_1
? ../2/in_b_2
? ../in_b
$ hg status --cwd b/2
? a/1/in_a_1
? a/in_a
? b/1/in_b_1
? b/2/in_b_2
? b/in_b
? in_root
$ hg status --cwd b/2 .
? in_b_2
$ hg status --cwd b/2 ..
? ../1/in_b_1
? in_b_2
? ../in_b
combining patterns with root and patterns without a root works
$ hg st a/in_a re:.*b$
? a/in_a
? b/in_b
$ cd ..
$ hg init repo2
$ cd repo2
$ touch modified removed deleted ignored
$ echo "^ignored$" > .hgignore
$ hg ci -A -m 'initial checkin'
adding .hgignore
adding deleted
adding modified
adding removed
$ touch modified added unknown ignored
$ hg add added
$ hg remove removed
$ rm deleted
hg status:
$ hg status
A added
R removed
! deleted
? unknown
hg status modified added removed deleted unknown never-existed ignored:
$ hg status modified added removed deleted unknown never-existed ignored
never-existed: * (glob)
A added
R removed
! deleted
? unknown
$ hg copy modified copied
hg status -C:
$ hg status -C
A added
A copied
modified
R removed
! deleted
? unknown
hg status -A:
$ hg status -A
A added
A copied
modified
R removed
! deleted
? unknown
I ignored
C .hgignore
C modified
$ hg status -A -Tjson
[
{
"path": "added",
"status": "A"
},
{
"copy": "modified",
"path": "copied",
"status": "A"
},
{
"path": "removed",
"status": "R"
},
{
"path": "deleted",
"status": "!"
},
{
"path": "unknown",
"status": "?"
},
{
"path": "ignored",
"status": "I"
},
{
"path": ".hgignore",
"status": "C"
},
{
"path": "modified",
"status": "C"
}
]
$ hg status -A -Tpickle > pickle
>>> import pickle
>>> print sorted((x['status'], x['path']) for x in pickle.load(open("pickle")))
[('!', 'deleted'), ('?', 'pickle'), ('?', 'unknown'), ('A', 'added'), ('A', 'copied'), ('C', '.hgignore'), ('C', 'modified'), ('I', 'ignored'), ('R', 'removed')]
$ rm pickle
$ echo "^ignoreddir$" > .hgignore
$ mkdir ignoreddir
$ touch ignoreddir/file
Test templater support:
$ hg status -AT "[{status}]\t{if(copy, '{copy} -> ')}{path}\n"
[M] .hgignore
[A] added
[A] modified -> copied
[R] removed
[!] deleted
[?] ignored
[?] unknown
[I] ignoreddir/file
[C] modified
$ hg status -AT default
M .hgignore
A added
A copied
modified
R removed
! deleted
? ignored
? unknown
I ignoreddir/file
C modified
$ hg status -T compact
abort: "status" not in template map
[255]
hg status ignoreddir/file:
$ hg status ignoreddir/file
hg status -i ignoreddir/file:
$ hg status -i ignoreddir/file
I ignoreddir/file
$ cd ..
Check 'status -q' and some combinations
$ hg init repo3
$ cd repo3
$ touch modified removed deleted ignored
$ echo "^ignored$" > .hgignore
$ hg commit -A -m 'initial checkin'
adding .hgignore
adding deleted
adding modified
adding removed
$ touch added unknown ignored
$ hg add added
$ echo "test" >> modified
$ hg remove removed
$ rm deleted
$ hg copy modified copied
Specify working directory revision explicitly, that should be the same as
"hg status"
$ hg status --change "wdir()"
M modified
A added
A copied
R removed
! deleted
? unknown
Run status with 2 different flags.
Check if result is the same or different.
If result is not as expected, raise error
$ assert() {
> hg status $1 > ../a
> hg status $2 > ../b
> if diff ../a ../b > /dev/null; then
> out=0
> else
> out=1
> fi
> if [ $3 -eq 0 ]; then
> df="same"
> else
> df="different"
> fi
> if [ $out -ne $3 ]; then
> echo "Error on $1 and $2, should be $df."
> fi
> }
Assert flag1 flag2 [0-same | 1-different]
$ assert "-q" "-mard" 0
$ assert "-A" "-marduicC" 0
$ assert "-qA" "-mardcC" 0
$ assert "-qAui" "-A" 0
$ assert "-qAu" "-marducC" 0
$ assert "-qAi" "-mardicC" 0
$ assert "-qu" "-u" 0
$ assert "-q" "-u" 1
$ assert "-m" "-a" 1
$ assert "-r" "-d" 1
$ cd ..
$ hg init repo4
$ cd repo4
$ touch modified removed deleted
$ hg ci -q -A -m 'initial checkin'
$ touch added unknown
$ hg add added
$ hg remove removed
$ rm deleted
$ echo x > modified
$ hg copy modified copied
$ hg ci -m 'test checkin' -d "1000001 0"
$ rm *
$ touch unrelated
$ hg ci -q -A -m 'unrelated checkin' -d "1000002 0"
hg status --change 1:
$ hg status --change 1
M modified
A added
A copied
R removed
hg status --change 1 unrelated:
$ hg status --change 1 unrelated
hg status -C --change 1 added modified copied removed deleted:
$ hg status -C --change 1 added modified copied removed deleted
M modified
A added
A copied
modified
R removed
hg status -A --change 1 and revset:
$ hg status -A --change '1|1'
M modified
A added
A copied
modified
R removed
C deleted
$ cd ..
hg status of binary file starting with '\1\n', a separator for metadata:
$ hg init repo5
$ cd repo5
>>> open("010a", "wb").write("\1\nfoo")
$ hg ci -q -A -m 'initial checkin'
$ hg status -A
C 010a
>>> open("010a", "wb").write("\1\nbar")
$ hg status -A
M 010a
$ hg ci -q -m 'modify 010a'
$ hg status -A --rev 0:1
M 010a
$ touch empty
$ hg ci -q -A -m 'add another file'
$ hg status -A --rev 1:2 010a
C 010a
$ cd ..
test "hg status" with "directory pattern" which matches against files
only known on target revision.
$ hg init repo6
$ cd repo6
$ echo a > a.txt
$ hg add a.txt
$ hg commit -m '#0'
$ mkdir -p 1/2/3/4/5
$ echo b > 1/2/3/4/5/b.txt
$ hg add 1/2/3/4/5/b.txt
$ hg commit -m '#1'
$ hg update -C 0 > /dev/null
$ hg status -A
C a.txt
the directory matching against specified pattern should be removed,
because directory existence prevents 'dirstate.walk()' from showing
warning message about such pattern.
$ test ! -d 1
$ hg status -A --rev 1 1/2/3/4/5/b.txt
R 1/2/3/4/5/b.txt
$ hg status -A --rev 1 1/2/3/4/5
R 1/2/3/4/5/b.txt
$ hg status -A --rev 1 1/2/3
R 1/2/3/4/5/b.txt
$ hg status -A --rev 1 1
R 1/2/3/4/5/b.txt
$ hg status --config ui.formatdebug=True --rev 1 1
status = [
{*'path': '1/2/3/4/5/b.txt'*}, (glob)
]
#if windows
$ hg --config ui.slash=false status -A --rev 1 1
R 1\2\3\4\5\b.txt
#endif
$ cd ..
Status after move overwriting a file (issue4458)
=================================================
$ hg init issue4458
$ cd issue4458
$ echo a > a
$ echo b > b
$ hg commit -Am base
adding a
adding b
with --force
$ hg mv b --force a
$ hg st --copies
M a
b
R b
$ hg revert --all
reverting a
undeleting b
$ rm *.orig
without force
$ hg rm a
$ hg st --copies
R a
$ hg mv b a
$ hg st --copies
M a
b
R b
using ui.statuscopies setting
$ hg st --config ui.statuscopies=true
M a
b
R b
$ hg st --config ui.statuscopies=false
M a
R b
Other "bug" highlight, the revision status does not report the copy information.
This is buggy behavior.
$ hg commit -m 'blah'
$ hg st --copies --change .
M a
R b
$ cd ..