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
$ cat >> fakepager.py <<EOF
> import sys
> printed = False
> for line in sys.stdin:
> sys.stdout.write('paged! %r\n' % line)
> printed = True
> if not printed:
> sys.stdout.write('paged empty output!\n')
> EOF
Enable ui.formatted because pager won't fire without it, and set up
pager and tell it to use our fake pager that lets us see when the
pager was running.
$ cat >> $HGRCPATH <<EOF
> [ui]
> formatted = yes
> color = no
> [pager]
> pager = "$PYTHON" $TESTTMP/fakepager.py
> EOF
$ hg init repo
$ cd repo
$ echo a >> a
$ hg add a
$ hg ci -m 'add a'
$ for x in `"$PYTHON" $TESTDIR/seq.py 1 10`; do
> echo a $x >> a
> hg ci -m "modify a $x"
> done
By default diff and log are paged, but id is not:
$ hg diff -c 2 --pager=yes
paged! 'diff -r f4be7687d414 -r bce265549556 a\n'
paged! '--- a/a\tThu Jan 01 00:00:00 1970 +0000\n'
paged! '+++ b/a\tThu Jan 01 00:00:00 1970 +0000\n'
paged! '@@ -1,2 +1,3 @@\n'
paged! ' a\n'
paged! ' a 1\n'
paged! '+a 2\n'
$ hg log --limit 2
paged! 'changeset: 10:46106edeeb38\n'
paged! 'tag: tip\n'
paged! 'user: test\n'
paged! 'date: Thu Jan 01 00:00:00 1970 +0000\n'
paged! 'summary: modify a 10\n'
paged! '\n'
paged! 'changeset: 9:6dd8ea7dd621\n'
paged! 'user: test\n'
paged! 'date: Thu Jan 01 00:00:00 1970 +0000\n'
paged! 'summary: modify a 9\n'
paged! '\n'
$ hg id
46106edeeb38 tip
We can control the pager from the config
$ hg log --limit 1 --config 'ui.paginate=False'
changeset: 10:46106edeeb38
tag: tip
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: modify a 10
$ hg log --limit 1 --config 'ui.paginate=0'
changeset: 10:46106edeeb38
tag: tip
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: modify a 10
$ hg log --limit 1 --config 'ui.paginate=1'
paged! 'changeset: 10:46106edeeb38\n'
paged! 'tag: tip\n'
paged! 'user: test\n'
paged! 'date: Thu Jan 01 00:00:00 1970 +0000\n'
paged! 'summary: modify a 10\n'
paged! '\n'
explicit --pager=on should take precedence over other configurations
(issue5580)
$ cat >> $HGRCPATH <<EOF
> [ui]
> paginate = false
> EOF
$ hg log --limit 1 --pager=on
paged! 'changeset: 10:46106edeeb38\n'
paged! 'tag: tip\n'
paged! 'user: test\n'
paged! 'date: Thu Jan 01 00:00:00 1970 +0000\n'
paged! 'summary: modify a 10\n'
paged! '\n'
$ cat >> $HGRCPATH <<EOF
> [ui]
> # true is default value of ui.paginate
> paginate = true
> EOF
$ hg log --limit 1 --pager=off
changeset: 10:46106edeeb38
tag: tip
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: modify a 10
We can enable the pager on id:
BROKEN: should be paged
$ hg --config pager.attend-id=yes id
46106edeeb38 tip
Setting attend-$COMMAND to a false value works, even with pager in
core:
$ hg --config pager.attend-diff=no diff -c 2
diff -r f4be7687d414 -r bce265549556 a
--- a/a Thu Jan 01 00:00:00 1970 +0000
+++ b/a Thu Jan 01 00:00:00 1970 +0000
@@ -1,2 +1,3 @@
a
a 1
+a 2
Command aliases should have same behavior as main command
$ hg history --limit 2
paged! 'changeset: 10:46106edeeb38\n'
paged! 'tag: tip\n'
paged! 'user: test\n'
paged! 'date: Thu Jan 01 00:00:00 1970 +0000\n'
paged! 'summary: modify a 10\n'
paged! '\n'
paged! 'changeset: 9:6dd8ea7dd621\n'
paged! 'user: test\n'
paged! 'date: Thu Jan 01 00:00:00 1970 +0000\n'
paged! 'summary: modify a 9\n'
paged! '\n'
Abbreviated command alias should also be paged
$ hg hist -l 1
paged! 'changeset: 10:46106edeeb38\n'
paged! 'tag: tip\n'
paged! 'user: test\n'
paged! 'date: Thu Jan 01 00:00:00 1970 +0000\n'
paged! 'summary: modify a 10\n'
paged! '\n'
Attend for an abbreviated command does not work
$ hg --config pager.attend-ident=true ident
46106edeeb38 tip
$ hg --config extensions.pager= --config pager.attend-ident=true ident
46106edeeb38 tip
Pager should not start if stdout is not a tty.
$ hg log -l1 -q --config ui.formatted=False
10:46106edeeb38
Pager should be disabled if pager.pager is empty (otherwise the output would
be silently lost.)
$ hg log -l1 -q --config pager.pager=
10:46106edeeb38
Pager with color enabled allows colors to come through by default,
even though stdout is no longer a tty.
$ cat >> $HGRCPATH <<EOF
> [ui]
> color = always
> [color]
> mode = ansi
> EOF
$ hg log --limit 3
paged! '\x1b[0;33mchangeset: 10:46106edeeb38\x1b[0m\n'
paged! 'tag: tip\n'
paged! 'user: test\n'
paged! 'date: Thu Jan 01 00:00:00 1970 +0000\n'
paged! 'summary: modify a 10\n'
paged! '\n'
paged! '\x1b[0;33mchangeset: 9:6dd8ea7dd621\x1b[0m\n'
paged! 'user: test\n'
paged! 'date: Thu Jan 01 00:00:00 1970 +0000\n'
paged! 'summary: modify a 9\n'
paged! '\n'
paged! '\x1b[0;33mchangeset: 8:cff05a6312fe\x1b[0m\n'
paged! 'user: test\n'
paged! 'date: Thu Jan 01 00:00:00 1970 +0000\n'
paged! 'summary: modify a 8\n'
paged! '\n'
#if no-chg
An invalid pager command name is reported sensibly if we don't have to
use shell=True in the subprocess call:
$ hg log --limit 3 --config pager.pager=this-command-better-never-exist
missing pager command 'this-command-better-never-exist', skipping pager
\x1b[0;33mchangeset: 10:46106edeeb38\x1b[0m (esc)
tag: tip
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: modify a 10
\x1b[0;33mchangeset: 9:6dd8ea7dd621\x1b[0m (esc)
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: modify a 9
\x1b[0;33mchangeset: 8:cff05a6312fe\x1b[0m (esc)
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: modify a 8
#endif
A complicated pager command gets worse behavior. Bonus points if you can
improve this.
$ hg log --limit 3 \
> --config pager.pager='this-command-better-never-exist --seriously' \
> 2>/dev/null || true
Pager works with shell aliases.
$ cat >> $HGRCPATH <<EOF
> [alias]
> echoa = !echo a
> EOF
$ hg echoa
a
BROKEN: should be paged
$ hg --config pager.attend-echoa=yes echoa
a
Pager works with hg aliases including environment variables.
$ cat >> $HGRCPATH <<'EOF'
> [alias]
> printa = log -T "$A\n" -r 0
> EOF
$ A=1 hg --config pager.attend-printa=yes printa
paged! '1\n'
$ A=2 hg --config pager.attend-printa=yes printa
paged! '2\n'
Something that's explicitly attended is still not paginated if the
pager is globally set to off using a flag:
$ A=2 hg --config pager.attend-printa=yes printa --pager=no
2
Pager should not override the exit code of other commands
$ cat >> $TESTTMP/fortytwo.py <<'EOF'
> from mercurial import commands, registrar
> cmdtable = {}
> command = registrar.command(cmdtable)
> @command(b'fortytwo', [], b'fortytwo', norepo=True)
> def fortytwo(ui, *opts):
> ui.write(b'42\n')
> return 42
> EOF
$ cat >> $HGRCPATH <<'EOF'
> [extensions]
> fortytwo = $TESTTMP/fortytwo.py
> EOF
$ hg fortytwo --pager=on
paged! '42\n'
[42]
A command that asks for paging using ui.pager() directly works:
$ hg blame a
paged! ' 0: a\n'
paged! ' 1: a 1\n'
paged! ' 2: a 2\n'
paged! ' 3: a 3\n'
paged! ' 4: a 4\n'
paged! ' 5: a 5\n'
paged! ' 6: a 6\n'
paged! ' 7: a 7\n'
paged! ' 8: a 8\n'
paged! ' 9: a 9\n'
paged! '10: a 10\n'
but not with HGPLAIN
$ HGPLAIN=1 hg blame a
0: a
1: a 1
2: a 2
3: a 3
4: a 4
5: a 5
6: a 6
7: a 7
8: a 8
9: a 9
10: a 10
explicit flags work too:
$ hg blame --pager=no a
0: a
1: a 1
2: a 2
3: a 3
4: a 4
5: a 5
6: a 6
7: a 7
8: a 8
9: a 9
10: a 10
A command with --output option:
$ hg cat -r0 a
paged! 'a\n'
$ hg cat -r0 a --output=-
paged! 'a\n'
$ hg cat -r0 a --output=out
$ hg export -r0
paged! '# HG changeset patch\n'
paged! '# User test\n'
paged! '# Date 0 0\n'
paged! '# Thu Jan 01 00:00:00 1970 +0000\n'
paged! '# Node ID 1f0dee641bb7258c56bd60e93edfa2405381c41e\n'
paged! '# Parent 0000000000000000000000000000000000000000\n'
paged! 'add a\n'
paged! '\n'
paged! '\x1b[0;1mdiff -r 000000000000 -r 1f0dee641bb7 a\x1b[0m\n'
paged! '\x1b[0;31;1m--- /dev/null\tThu Jan 01 00:00:00 1970 +0000\x1b[0m\n'
paged! '\x1b[0;32;1m+++ b/a\tThu Jan 01 00:00:00 1970 +0000\x1b[0m\n'
paged! '\x1b[0;35m@@ -0,0 +1,1 @@\x1b[0m\n'
paged! '\x1b[0;32m+a\x1b[0m\n'
$ hg export -r0 -o -
paged! '# HG changeset patch\n'
paged! '# User test\n'
paged! '# Date 0 0\n'
paged! '# Thu Jan 01 00:00:00 1970 +0000\n'
paged! '# Node ID 1f0dee641bb7258c56bd60e93edfa2405381c41e\n'
paged! '# Parent 0000000000000000000000000000000000000000\n'
paged! 'add a\n'
paged! '\n'
paged! '\x1b[0;1mdiff -r 000000000000 -r 1f0dee641bb7 a\x1b[0m\n'
paged! '\x1b[0;31;1m--- /dev/null\tThu Jan 01 00:00:00 1970 +0000\x1b[0m\n'
paged! '\x1b[0;32;1m+++ b/a\tThu Jan 01 00:00:00 1970 +0000\x1b[0m\n'
paged! '\x1b[0;35m@@ -0,0 +1,1 @@\x1b[0m\n'
paged! '\x1b[0;32m+a\x1b[0m\n'
$ hg export -r0 -o out
$ rm out
Put annotate in the ignore list for pager:
$ cat >> $HGRCPATH <<EOF
> [pager]
> ignore = annotate
> EOF
$ hg blame a
0: a
1: a 1
2: a 2
3: a 3
4: a 4
5: a 5
6: a 6
7: a 7
8: a 8
9: a 9
10: a 10
During pushbuffer, pager should not start:
$ cat > $TESTTMP/pushbufferpager.py <<EOF
> def uisetup(ui):
> ui.pushbuffer()
> ui.pager(b'mycmd')
> ui.write(b'content\n')
> ui.write(ui.popbuffer())
> EOF
$ echo append >> a
$ hg --config extensions.pushbuffer=$TESTTMP/pushbufferpager.py status --color=off
content
paged! 'M a\n'
Environment variables like LESS and LV are set automatically:
$ cat > $TESTTMP/printlesslv.py <<EOF
> from __future__ import absolute_import
> import os
> import sys
> sys.stdin.read()
> for name in ['LESS', 'LV']:
> sys.stdout.write(('%s=%s\n') % (name, os.environ.get(name, '-')))
> sys.stdout.flush()
> EOF
$ cat >> $HGRCPATH <<EOF
> [alias]
> noop = log -r 0 -T ''
> [ui]
> formatted=1
> [pager]
> pager = "$PYTHON" $TESTTMP/printlesslv.py
> EOF
$ unset LESS
$ unset LV
$ hg noop --pager=on
LESS=FRX
LV=-c
$ LESS=EFGH hg noop --pager=on
LESS=EFGH
LV=-c