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
$ hg init
no bookmarks
$ hg bookmarks
no bookmarks set
set bookmark X
$ hg bookmark X
list bookmarks
$ hg bookmark
* X -1:000000000000
list bookmarks with color
$ hg --config extensions.color= --config color.mode=ansi \
> bookmark --color=always
\x1b[0;32m * \x1b[0m\x1b[0;32mX\x1b[0m\x1b[0;32m -1:000000000000\x1b[0m (esc)
update to bookmark X
$ hg bookmarks
* X -1:000000000000
$ hg update X
0 files updated, 0 files merged, 0 files removed, 0 files unresolved
list bookmarks
$ hg bookmarks
* X -1:000000000000
rename
$ hg bookmark -m X Z
list bookmarks
$ cat .hg/bookmarks.current
Z (no-eol)
$ cat .hg/bookmarks
0000000000000000000000000000000000000000 Z
$ hg bookmarks
* Z -1:000000000000
new bookmarks X and Y, first one made active
$ hg bookmark Y X
list bookmarks
$ hg bookmark
X -1:000000000000
* Y -1:000000000000
Z -1:000000000000
$ hg bookmark -d X
commit
$ echo 'b' > b
$ hg add b
$ hg commit -m'test'
list bookmarks
$ hg bookmark
* Y 0:719295282060
Z -1:000000000000
Verify that switching to Z updates the active bookmark:
$ hg update Z
0 files updated, 0 files merged, 1 files removed, 0 files unresolved
(activating bookmark Z)
$ hg bookmark
Y 0:719295282060
* Z -1:000000000000
Switch back to Y for the remaining tests in this file:
$ hg update Y
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
(activating bookmark Y)
delete bookmarks
$ hg bookmark -d Y
$ hg bookmark -d Z
list bookmarks
$ hg bookmark
no bookmarks set
update to tip
$ hg update tip
0 files updated, 0 files merged, 0 files removed, 0 files unresolved
set bookmark Y using -r . but make sure that the active
bookmark is not activated
$ hg bookmark -r . Y
list bookmarks, Y should not be active
$ hg bookmark
Y 0:719295282060
now, activate Y
$ hg up -q Y
set bookmark Z using -i
$ hg bookmark -r . -i Z
$ hg bookmarks
* Y 0:719295282060
Z 0:719295282060
deactivate active bookmark using -i
$ hg bookmark -i Y
$ hg bookmarks
Y 0:719295282060
Z 0:719295282060
$ hg up -q Y
$ hg bookmark -i
$ hg bookmarks
Y 0:719295282060
Z 0:719295282060
$ hg bookmark -i
no active bookmark
$ hg up -q Y
$ hg bookmarks
* Y 0:719295282060
Z 0:719295282060
deactivate active bookmark while renaming
$ hg bookmark -i -m Y X
$ hg bookmarks
X 0:719295282060
Z 0:719295282060
bare update moves the active bookmark forward and clear the divergent bookmarks
$ echo a > a
$ hg ci -Am1
adding a
$ echo b >> a
$ hg ci -Am2
$ hg bookmark X@1 -r 1
$ hg bookmark X@2 -r 2
$ hg update X
0 files updated, 0 files merged, 1 files removed, 0 files unresolved
(activating bookmark X)
$ hg bookmarks
* X 0:719295282060
X@1 1:cc586d725fbe
X@2 2:49e1c4e84c58
Z 0:719295282060
$ hg update
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
updating bookmark X
$ hg bookmarks
* X 2:49e1c4e84c58
Z 0:719295282060
test deleting .hg/bookmarks.current when explicitly updating
to a revision
$ echo a >> b
$ hg ci -m.
$ hg up -q X
$ test -f .hg/bookmarks.current
try to update to it again to make sure we don't
set and then unset it
$ hg up -q X
$ test -f .hg/bookmarks.current
$ hg up -q 1
$ test -f .hg/bookmarks.current
[1]
when a bookmark is active, hg up -r . is
analogous to hg book -i <active bookmark>
$ hg up -q X
$ hg up -q .
$ test -f .hg/bookmarks.current
[1]
issue 4552 -- simulate a pull moving the active bookmark
$ hg up -q X
$ printf "Z" > .hg/bookmarks.current
$ hg log -T '{activebookmark}\n' -r Z
Z
$ hg log -T '{bookmarks % "{active}\n"}' -r Z
Z
test that updating to closed branch head also advances active bookmark
$ hg commit --close-branch -m "closed"
$ hg update -q ".^1"
$ hg bookmark Y
$ hg bookmarks
X 3:4d6bd4bfb1ae
* Y 3:4d6bd4bfb1ae
Z 0:719295282060
$ hg update
0 files updated, 0 files merged, 0 files removed, 0 files unresolved
updating bookmark Y
$ hg bookmarks
X 3:4d6bd4bfb1ae
* Y 4:8fa964221e8e
Z 0:719295282060
$ hg parents -q
4:8fa964221e8e
Checks command to retrieve active bookmark
------------------------------------------
display how "{activebookmark}" template is unsuitable for the task
$ hg book -T '- {activebookmark}\n'
-
- Y
-
$ hg book -r . W
$ hg book -T '- {activebookmark}\n'
- Y
-
- Y
-
$ hg bookmarks -ql .
Y
$ hg bookmarks --inactive
$ hg bookmarks -ql .
abort: no active bookmark!
[255]