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
#testcases tree flat
$ . "$TESTDIR/narrow-library.sh"
#if tree
$ cat << EOF >> $HGRCPATH
> [experimental]
> treemanifest = 1
> EOF
#endif
$ hg init master
$ cd master
$ mkdir inside
$ echo 'inside' > inside/f
$ hg add inside/f
$ hg commit -m 'add inside'
$ mkdir widest
$ echo 'widest' > widest/f
$ hg add widest/f
$ hg commit -m 'add widest'
$ mkdir outside
$ echo 'outside' > outside/f
$ hg add outside/f
$ hg commit -m 'add outside'
$ cd ..
narrow clone the inside file
$ hg clone --narrow ssh://user@dummy/master narrow
requesting all changes
adding changesets
adding manifests
adding file changes
added 3 changesets with 0 changes to 0 files
new changesets *:* (glob)
updating to branch default
0 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ cd narrow
$ hg tracked
$ hg files
[1]
widen from an empty clone
$ hg tracked --addinclude inside
comparing with ssh://user@dummy/master
searching for changes
adding changesets
adding manifests
adding file changes
added 0 changesets with 1 changes to 1 files
$ hg tracked
I path:inside
$ ls
inside
$ cat inside/f
inside
$ cd ..
add more upstream files which we will include in a wider narrow spec
$ cd master
$ mkdir wider
$ echo 'wider' > wider/f
$ hg add wider/f
$ echo 'widest v2' > widest/f
$ hg commit -m 'add wider, update widest'
$ echo 'widest v3' > widest/f
$ hg commit -m 'update widest v3'
$ echo 'inside v2' > inside/f
$ hg commit -m 'update inside'
$ mkdir outside2
$ echo 'outside2' > outside2/f
$ hg add outside2/f
$ hg commit -m 'add outside2'
$ echo 'widest v4' > widest/f
$ hg commit -m 'update widest v4'
$ hg log -T "{if(ellipsis, '...')}{rev}: {desc}\n"
7: update widest v4
6: add outside2
5: update inside
4: update widest v3
3: add wider, update widest
2: add outside
1: add widest
0: add inside
$ cd ..
Widen the narrow spec to see the widest file. This should not get the newly
added upstream revisions.
$ cd narrow
$ hg id -n
2
$ hg tracked --addinclude widest/f --debug
comparing with ssh://user@dummy/master
running python "*dummyssh" *user@dummy* *hg -R master serve --stdio* (glob)
sending hello command
sending between command
remote: * (glob)
remote: capabilities: * (glob)
remote: 1
sending protocaps command
query 1; heads
sending batch command
searching for changes
all local heads known remotely
sending narrow_widen command
bundle2-input-bundle: with-transaction
bundle2-input-part: "changegroup" (params: * mandatory) supported (glob)
adding changesets
adding manifests
adding widest/ revisions (tree !)
adding file changes
adding widest/f revisions (tree !)
added 0 changesets with 1 changes to 1 files
bundle2-input-part: total payload size * (glob)
bundle2-input-bundle: 0 parts total
widest/f: add from widened narrow clone -> g
getting widest/f
$ hg tracked
I path:inside
I path:widest/f
$ cat widest/f
widest
$ hg id -n
2
Test that extending already included files should not call narrow_widen
wireprotocol command
$ hg tracked --addinclude widest/f
nothing to widen or narrow
Pull down the newly added upstream revision.
$ hg pull
pulling from ssh://user@dummy/master
searching for changes
adding changesets
adding manifests
adding file changes
added 5 changesets with 4 changes to 2 files
new changesets *:* (glob)
(run 'hg update' to get a working copy)
$ hg update -r 'desc("add wider")'
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ cat widest/f
widest v2
$ hg update -r 'desc("update inside")'
2 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ cat widest/f
widest v3
$ cat inside/f
inside v2
$ hg log -T "{if(ellipsis, '...')}{rev}: {desc}\n"
7: update widest v4
6: add outside2
5: update inside
4: update widest v3
3: add wider, update widest
2: add outside
1: add widest
0: add inside
Check that widening with a newline fails
$ hg tracked --addinclude 'widest
> '
abort: newlines are not allowed in narrowspec paths
[255]
widen the narrow spec to include the wider file
$ hg tracked --addinclude wider
comparing with ssh://user@dummy/master
searching for changes
adding changesets
adding manifests
adding file changes
added 0 changesets with 1 changes to 1 files
$ hg tracked
I path:inside
I path:wider
I path:widest/f
$ hg update 'desc("add widest")'
2 files updated, 0 files merged, 1 files removed, 0 files unresolved
$ cat widest/f
widest
$ hg update 'desc("add wider, update widest")'
2 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ cat wider/f
wider
$ cat widest/f
widest v2
$ hg update 'desc("update widest v3")'
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ cat widest/f
widest v3
$ hg update 'desc("update widest v4")'
2 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ cat widest/f
widest v4
$ hg log -T "{if(ellipsis, '...')}{rev}: {desc}\n"
7: update widest v4
6: add outside2
5: update inside
4: update widest v3
3: add wider, update widest
2: add outside
1: add widest
0: add inside
separate suite of tests: files from 0-10 modified in changes 0-10. This allows
more obvious precise tests tickling particular corner cases.
$ cd ..
$ hg init upstream
$ cd upstream
$ for x in `$TESTDIR/seq.py 0 10`
> do
> mkdir d$x
> echo $x > d$x/f
> hg add d$x/f
> hg commit -m "add d$x/f"
> done
$ hg log -T "{rev}: {desc}\n"
10: add d10/f
9: add d9/f
8: add d8/f
7: add d7/f
6: add d6/f
5: add d5/f
4: add d4/f
3: add d3/f
2: add d2/f
1: add d1/f
0: add d0/f
make narrow clone with every third node.
$ cd ..
$ hg clone --narrow ssh://user@dummy/upstream narrow2 --include d0 --include d3 --include d6 --include d9
requesting all changes
adding changesets
adding manifests
adding file changes
added 11 changesets with 4 changes to 4 files
new changesets *:* (glob)
updating to branch default
4 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ cd narrow2
$ hg tracked
I path:d0
I path:d3
I path:d6
I path:d9
$ hg verify
checking changesets
checking manifests
checking directory manifests (tree !)
crosschecking files in changesets and manifests
checking files
checked 11 changesets with 4 changes to 4 files
$ hg log -T "{if(ellipsis, '...')}{rev}: {desc}\n"
10: add d10/f
9: add d9/f
8: add d8/f
7: add d7/f
6: add d6/f
5: add d5/f
4: add d4/f
3: add d3/f
2: add d2/f
1: add d1/f
0: add d0/f
$ hg tracked --addinclude d1
comparing with ssh://user@dummy/upstream
searching for changes
adding changesets
adding manifests
adding file changes
added 0 changesets with 1 changes to 1 files
$ hg tracked
I path:d0
I path:d1
I path:d3
I path:d6
I path:d9
$ hg log -T "{if(ellipsis, '...')}{rev}: {desc}\n"
10: add d10/f
9: add d9/f
8: add d8/f
7: add d7/f
6: add d6/f
5: add d5/f
4: add d4/f
3: add d3/f
2: add d2/f
1: add d1/f
0: add d0/f
Verify shouldn't claim the repo is corrupt after a widen.
$ hg verify
checking changesets
checking manifests
checking directory manifests (tree !)
crosschecking files in changesets and manifests
checking files
checked 11 changesets with 5 changes to 5 files
Widening preserves parent of local commit
$ cd ..
$ hg clone -q --narrow ssh://user@dummy/upstream narrow3 --include d2 -r 2
$ cd narrow3
$ hg log -T "{if(ellipsis, '...')}{rev}: {desc}\n"
2: add d2/f
1: add d1/f
0: add d0/f
$ hg pull -q -r 3
$ hg co -q tip
$ hg pull -q -r 4
$ echo local > d2/f
$ hg ci -m local
created new head
$ hg tracked -q --addinclude d0 --addinclude d9
Widening preserves bookmarks
$ cd ..
$ hg clone -q --narrow ssh://user@dummy/upstream narrow-bookmarks --include d4
$ cd narrow-bookmarks
$ echo local > d4/f
$ hg ci -m local
$ hg bookmarks bookmark
$ hg bookmarks
* bookmark 11:* (glob)
$ hg -q tracked --addinclude d2
$ hg bookmarks
* bookmark 11:* (glob)
$ hg log -r bookmark -T '{desc}\n'
local
Widening that fails can be recovered from
$ cd ..
$ hg clone -q --narrow ssh://user@dummy/upstream interrupted --include d0
$ cd interrupted
$ echo local > d0/f
$ hg ci -m local
$ hg log -T "{if(ellipsis, '...')}{rev}: {desc}\n"
11: local
10: add d10/f
9: add d9/f
8: add d8/f
7: add d7/f
6: add d6/f
5: add d5/f
4: add d4/f
3: add d3/f
2: add d2/f
1: add d1/f
0: add d0/f
$ hg bookmarks bookmark
$ hg --config hooks.pretxnchangegroup.bad=false tracked --addinclude d1
comparing with ssh://user@dummy/upstream
searching for changes
adding changesets
adding manifests
adding file changes
added 0 changesets with 1 changes to 1 files
$ hg log -T "{if(ellipsis, '...')}{rev}: {desc}\n"
11: local
10: add d10/f
9: add d9/f
8: add d8/f
7: add d7/f
6: add d6/f
5: add d5/f
4: add d4/f
3: add d3/f
2: add d2/f
1: add d1/f
0: add d0/f
$ hg bookmarks
* bookmark 11:* (glob)
$ hg unbundle .hg/strip-backup/*-widen.hg
abort: .hg/strip-backup/*-widen.hg: $ENOTDIR$ (windows !)
abort: $ENOENT$: .hg/strip-backup/*-widen.hg (no-windows !)
[255]
$ hg log -T "{if(ellipsis, '...')}{rev}: {desc}\n"
11: local
10: add d10/f
9: add d9/f
8: add d8/f
7: add d7/f
6: add d6/f
5: add d5/f
4: add d4/f
3: add d3/f
2: add d2/f
1: add d1/f
0: add d0/f
$ hg bookmarks
* bookmark 11:* (glob)