revlog: automatically read from opened file handles
The revlog reading code commonly opens a new file handle for
reading on demand. There is support for passing a file handle
to revlog.revision(). But it is marked as an internal argument.
When revlogs are written, we write() data as it is available. But
we don't flush() data until all revisions are written.
Putting these two traits together, it is possible for an in-process
revlog reader during active writes to trigger the opening of a new
file handle on a file with unflushed writes. The reader won't have
access to all "available" revlog data (as it hasn't been flushed).
And with the introduction of the previous patch, this can lead to
the revlog raising an error due to a partial read.
I witnessed this behavior when applying changegroup data (via
`hg pull`) before issue6006 was fixed via different means. Having
this and the previous patch in play would have helped cause errors
earlier rather than manifesting as hash verification failures.
While this has been a long-standing issue, I believe the relatively
new delta computation code has tickled it into being more common.
This is because the new delta computation code will compute deltas
in more scenarios. This can lead to revlog reading. While the delta
computation code is probably supposed to reuse file handles, it
appears it isn't doing so in all circumstances.
But the issue runs deeper than that. Theoretically, any code can
access revision data during revlog writes. It appears we were just
getting lucky that it wasn't. (The "add revision callback" passed to
addgroup() provides an avenue to do this.)
If I changed the revlog's behavior to not cache the full revision
text or to clear caches after revision insertion during addgroup(),
I was able to produce crashes 100% of the time when writing changelog
revisions. This is because changelog's add revision callback attempts
to resolve the revision data to access the changed files list. And
without the revision's fulltext being cached, we performed a revlog
read, which required opening a new file handle. This attempted to read
unflushed data, leading to a partial read and a crash.
This commit teaches the revlog to store the file handles used for
writing multiple revisions during addgroup(). It also teaches the
code for resolving a file handle when reading to use these handles,
if available. This ensures that *any* reads (regardless of their
source) use the active writing file handles, if available. These
file handles have access to the unflushed data because they wrote it.
This allows reads to complete without issue.
Differential Revision: https://phab.mercurial-scm.org/D5267
#require cvs112
This feature requires use of builtin cvsps!
$ echo "[extensions]" >> $HGRCPATH
$ echo "convert = " >> $HGRCPATH
create cvs repository with one project
$ mkdir cvsrepo
$ cd cvsrepo
$ CVSROOT=`pwd`
$ export CVSROOT
$ CVS_OPTIONS=-f
$ export CVS_OPTIONS
$ cd ..
$ rmdir cvsrepo
$ cvscall()
> {
> cvs -f "$@"
> }
output of 'cvs ci' varies unpredictably, so just discard it
$ cvsci()
> {
> sleep 1
> cvs -f ci "$@" >/dev/null
> }
$ cvscall -d "$CVSROOT" init
$ mkdir cvsrepo/proj
$ cvscall -q co proj
create file1 on the trunk
$ cd proj
$ touch file1
$ cvscall -Q add file1
$ cvsci -m"add file1 on trunk" file1
create two branches
$ cvscall -q tag -b v1_0
T file1
$ cvscall -q tag -b v1_1
T file1
create file2 on branch v1_0
$ cvscall -Q up -rv1_0
$ touch file2
$ cvscall -Q add file2
$ cvsci -m"add file2" file2
create file3, file4 on branch v1_1
$ cvscall -Q up -rv1_1
$ touch file3
$ touch file4
$ cvscall -Q add file3 file4
$ cvsci -m"add file3, file4 on branch v1_1" file3 file4
merge file2 from v1_0 to v1_1
$ cvscall -Q up -jv1_0
$ cvsci -m"MERGE from v1_0: add file2"
cvs commit: Examining .
Step things up a notch: now we make the history really hairy, with
changes bouncing back and forth between trunk and v1_2 and merges
going both ways. (I.e., try to model the real world.)
create branch v1_2
$ cvscall -Q up -A
$ cvscall -q tag -b v1_2
T file1
create file5 on branch v1_2
$ cvscall -Q up -rv1_2
$ touch file5
$ cvs -Q add file5
$ cvsci -m"add file5 on v1_2"
cvs commit: Examining .
create file6 on trunk post-v1_2
$ cvscall -Q up -A
$ touch file6
$ cvscall -Q add file6
$ cvsci -m"add file6 on trunk post-v1_2"
cvs commit: Examining .
merge file5 from v1_2 to trunk
$ cvscall -Q up -A
$ cvscall -Q up -jv1_2 file5
$ cvsci -m"MERGE from v1_2: add file5"
cvs commit: Examining .
merge file6 from trunk to v1_2
$ cvscall -Q up -rv1_2
$ cvscall up -jHEAD file6
U file6
$ cvsci -m"MERGE from HEAD: add file6"
cvs commit: Examining .
cvs rlog output
$ cvscall -q rlog proj | egrep '^(RCS file|revision)'
RCS file: $TESTTMP/cvsrepo/proj/file1,v
revision 1.1
RCS file: $TESTTMP/cvsrepo/proj/Attic/file2,v
revision 1.1
revision 1.1.4.2
revision 1.1.4.1
revision 1.1.2.1
RCS file: $TESTTMP/cvsrepo/proj/Attic/file3,v
revision 1.1
revision 1.1.2.1
RCS file: $TESTTMP/cvsrepo/proj/Attic/file4,v
revision 1.1
revision 1.1.2.1
RCS file: $TESTTMP/cvsrepo/proj/file5,v
revision 1.2
revision 1.1
revision 1.1.2.1
RCS file: $TESTTMP/cvsrepo/proj/file6,v
revision 1.1
revision 1.1.2.2
revision 1.1.2.1
convert to hg (#1)
$ cd ..
$ hg convert --datesort proj proj.hg
initializing destination proj.hg repository
connecting to $TESTTMP/cvsrepo
scanning source...
collecting CVS rlog
15 log entries
creating changesets
9 changeset entries
sorting...
converting...
8 add file1 on trunk
7 add file2
6 MERGE from v1_0: add file2
5 file file3 was initially added on branch v1_1.
4 add file3, file4 on branch v1_1
3 add file5 on v1_2
2 add file6 on trunk post-v1_2
1 MERGE from HEAD: add file6
0 MERGE from v1_2: add file5
hg log -G output (#1)
$ hg -R proj.hg log -G --template "{rev} {desc}\n"
o 8 MERGE from v1_2: add file5
|
| o 7 MERGE from HEAD: add file6
| |
o | 6 add file6 on trunk post-v1_2
| |
| o 5 add file5 on v1_2
| |
| | o 4 add file3, file4 on branch v1_1
| | |
o | | 3 file file3 was initially added on branch v1_1.
|/ /
| o 2 MERGE from v1_0: add file2
|/
| o 1 add file2
|/
o 0 add file1 on trunk
convert to hg (#2: with merge detection)
$ hg convert \
> --config convert.cvsps.mergefrom='"^MERGE from (\S+):"' \
> --datesort \
> proj proj.hg2
initializing destination proj.hg2 repository
connecting to $TESTTMP/cvsrepo
scanning source...
collecting CVS rlog
15 log entries
creating changesets
9 changeset entries
sorting...
converting...
8 add file1 on trunk
7 add file2
6 MERGE from v1_0: add file2
5 file file3 was initially added on branch v1_1.
4 add file3, file4 on branch v1_1
3 add file5 on v1_2
2 add file6 on trunk post-v1_2
1 MERGE from HEAD: add file6
0 MERGE from v1_2: add file5
hg log -G output (#2)
$ hg -R proj.hg2 log -G --template "{rev} {desc}\n"
o 8 MERGE from v1_2: add file5
|
| o 7 MERGE from HEAD: add file6
| |
o | 6 add file6 on trunk post-v1_2
| |
| o 5 add file5 on v1_2
| |
| | o 4 add file3, file4 on branch v1_1
| | |
o | | 3 file file3 was initially added on branch v1_1.
|/ /
| o 2 MERGE from v1_0: add file2
|/
| o 1 add file2
|/
o 0 add file1 on trunk