phases: rework the logic of _pushdiscoveryphase to bound complexity
This rework the various graph traversal in _pushdiscoveryphase to keep the
complexity in check.
This is done though a couple of things:
- first, limiting the space we have to explore, for example, if we are not in
publishing push, we don't need to consider remote draft roots that are also
draft locally, as there is nothing to be moved there.
- avoid unbounded descendant computation, and use the faster "rev between"
computation.
This provide a massive boost to performance when exchanging with repository with
a massive amount of draft, like mozilla-try:
### data-env-vars.name = mozilla-try-2023-03-22-zstd-sparse-revlog
# benchmark.name = hg.command.push
# bin-env-vars.hg.flavor = default
# bin-env-vars.hg.py-re2-module = default
# benchmark.variants.explicit-rev = all-out-heads
# benchmark.variants.issue6528 = disabled
# benchmark.variants.protocol = ssh
# benchmark.variants.reuse-external-delta-parent = default
## benchmark.variants.revs = any-1-extra-rev
before: 20.346590 seconds
after: 11.232059 seconds (-38.15%, -7.48 seconds)
## benchmark.variants.revs = any-100-extra-rev
before: 24.752051 seconds
after: 15.367412 seconds (-37.91%, -9.38 seconds)
After this changes, the push operation is still quite too slow. Some of this
can be attributed to general phases slowness (reading all the roots from disk
for example) and other know slowness (not using persistent-nodemap, branchmap,
tags, etc. We are also working on them, but with this series, phase discovery
during push no longer showing up in profile and this is a pretty nice and bit
low-hanging fruit out of the way.
### (same case as the above)
# benchmark.variants.revs = any-1-extra-rev
pre-%ln-change: 44.235070
this-changeset: 11.232059 seconds (-74.61%, -33.00 seconds)
# benchmark.variants.revs = any-100-extra-rev
pre-%ln-change: 49.234697
this-changeset: 15.367412 seconds (-68.79%, -33.87 seconds)
Note that with this change, the `hg push` performance is now much closer to the
`hg pull` performance, even it still lagging behind a bit. (and the overall
performance are still too slow).
### data-env-vars.name = mozilla-try-2023-03-22-ds2-pnm
# benchmark.variants.explicit-rev = all-out-heads
# benchmark.variants.issue6528 = disabled
# benchmark.variants.protocol = ssh
# benchmark.variants.pulled-delta-reuse-policy = default
# bin-env-vars.hg.flavor = rust
## benchmark.variants.revs = any-1-extra-rev
hg.command.pull: 6.517450
hg.command.push: 11.219888
## benchmark.variants.revs = any-100-extra-rev
hg.command.pull: 10.160991
hg.command.push: 14.251107
### data-env-vars.name = mozilla-try-2023-03-22-zstd-sparse-revlog
# bin-env-vars.hg.py-re2-module = default
# benchmark.variants.explicit-rev = all-out-heads
# benchmark.variants.issue6528 = disabled
# benchmark.variants.protocol = ssh
# benchmark.variants.pulled-delta-reuse-policy = default
## bin-env-vars.hg.flavor = default
## benchmark.variants.revs = any-1-extra-rev
hg.command.pull: 8.577772
hg.command.push: 11.232059
## bin-env-vars.hg.flavor = default
## benchmark.variants.revs = any-100-extra-rev
hg.command.pull: 13.152976
hg.command.push: 15.367412
## bin-env-vars.hg.flavor = rust
## benchmark.variants.revs = any-1-extra-rev
hg.command.pull: 8.731982
hg.command.push: 11.178751
## bin-env-vars.hg.flavor = rust
## benchmark.variants.revs = any-100-extra-rev
hg.command.pull: 13.184236
hg.command.push: 15.620843
$ cat >> $HGRCPATH <<EOF
> [extensions]
> rebase=
> [phases]
> publish=False
> [merge]
> EOF
$ hg init repo
$ cd repo
$ echo a > a
$ echo b > b
$ hg commit -qAm ab
$ echo c >> a
$ echo c >> b
$ hg commit -qAm c
$ hg up -q ".^"
$ echo d >> a
$ echo d >> b
$ hg commit -qAm d
Testing on-failure=continue
$ echo on-failure=continue >> $HGRCPATH
$ hg rebase -s 1 -d 2 --tool false
rebasing 1:1f28a51c3c9b "c"
merging a
merging a failed!
merging b
merging b failed!
unresolved conflicts (see 'hg resolve', then 'hg rebase --continue')
[240]
$ hg resolve --list
U a
U b
$ hg rebase --abort
rebase aborted
Testing on-failure=halt
$ echo on-failure=halt >> $HGRCPATH
$ hg rebase -s 1 -d 2 --tool false
rebasing 1:1f28a51c3c9b "c"
merging a
merging a failed!
unresolved conflicts (see 'hg resolve', then 'hg rebase --continue')
[240]
$ hg resolve --list
U a
U b
$ hg rebase --abort
rebase aborted
Testing on-failure=prompt
$ cat <<EOS >> $HGRCPATH
> [merge]
> on-failure=prompt
> [ui]
> interactive=1
> EOS
$ cat <<EOS | hg rebase -s 1 -d 2 --tool false
> y
> n
> EOS
rebasing 1:1f28a51c3c9b "c"
merging a
merging a failed!
continue merge operation (yn)? y
merging b
merging b failed!
continue merge operation (yn)? n
unresolved conflicts (see 'hg resolve', then 'hg rebase --continue')
[240]
$ hg resolve --list
U a
U b
$ hg rebase --abort
rebase aborted
Check that successful tool with failed post-check halts the merge
$ cat <<EOS >> $HGRCPATH
> [merge-tools]
> true.check=changed
> EOS
$ cat <<EOS | hg rebase -s 1 -d 2 --tool true
> y
> n
> n
> EOS
rebasing 1:1f28a51c3c9b "c"
merging a
output file a appears unchanged
was merge successful (yn)? y
merging b
output file b appears unchanged
was merge successful (yn)? n
merging b failed!
continue merge operation (yn)? n
unresolved conflicts (see 'hg resolve', then 'hg rebase --continue')
[240]
$ hg resolve --list
R a
U b
$ hg rebase --abort
rebase aborted
Check that conflicts with conflict check also halts the merge
$ cat <<EOS >> $HGRCPATH
> [merge-tools]
> true.check=conflicts
> true.premerge=keep
> [merge]
> on-failure=halt
> EOS
$ hg rebase -s 1 -d 2 --tool true
rebasing 1:1f28a51c3c9b "c"
merging a
merging a failed!
unresolved conflicts (see 'hg resolve', then 'hg rebase --continue')
[240]
$ hg resolve --list
U a
U b
$ hg rebase --abort
rebase aborted
Check that always-prompt also can halt the merge
$ cat <<EOS | hg rebase -s 1 -d 2 --tool true --config merge-tools.true.check=prompt
> y
> n
> EOS
rebasing 1:1f28a51c3c9b "c"
merging a
was merge of 'a' successful (yn)? y
merging b
was merge of 'b' successful (yn)? n
merging b failed!
unresolved conflicts (see 'hg resolve', then 'hg rebase --continue')
[240]
$ hg resolve --list
R a
U b
$ hg rebase --abort
rebase aborted
Check that successful tool otherwise allows the merge to continue
$ hg rebase -s 1 -d 2 --tool echo --keep --config merge-tools.echo.premerge=keep
rebasing 1:1f28a51c3c9b "c"
merging a
$TESTTMP/repo/a *a~base* *a~other* (glob)
merging b
$TESTTMP/repo/b *b~base* *b~other* (glob)
Check that unshelve isn't broken by halting the merge
$ cat <<EOS >> $HGRCPATH
> [extensions]
> shelve =
> [merge-tools]
> false.check=conflicts
> false.premerge=false
> EOS
$ echo foo > shelve_file1
$ echo foo > shelve_file2
$ hg ci -qAm foo
$ echo bar >> shelve_file1
$ echo bar >> shelve_file2
$ hg shelve --list
$ hg shelve
shelved as default
2 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ echo baz >> shelve_file1
$ echo baz >> shelve_file2
$ hg ci -m baz
$ hg unshelve --tool false --config merge-tools.false.premerge=keep
unshelving change 'default'
rebasing shelved changes
merging shelve_file1
merging shelve_file1 failed!
unresolved conflicts (see 'hg resolve', then 'hg unshelve --continue')
[240]
$ hg status --config commands.status.verbose=True
M shelve_file1
M shelve_file2
? shelve_file1.orig
# The repository is in an unfinished *unshelve* state.
# Unresolved merge conflicts:
#
# shelve_file1
# shelve_file2
#
# To mark files as resolved: hg resolve --mark FILE
# To continue: hg unshelve --continue
# To abort: hg unshelve --abort
$ hg resolve --tool false --all --re-merge
merging shelve_file1
merging shelve_file1 failed!
merge halted after failed merge (see hg resolve)
[240]
$ hg shelve --list
default (*s ago) * changes to: foo (glob)
$ hg unshelve --abort
unshelve of 'default' aborted