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
============================================
Tests for the admin::chainsaw-update command
============================================
setup
=====
$ hg init src
$ cd src
$ echo 1 > root
$ hg add root
$ hg ci -Am R_0
$ hg branch A
marked working directory as branch A
(branches are permanent and global, did you want a bookmark?)
$ echo 42 > bar
$ hg add bar
$ hg ci -Am A_0
$ echo 1337 > bar
$ hg ci -Am A_1
$ hg update 'desc(R_0)'
0 files updated, 0 files merged, 1 files removed, 0 files unresolved
$ echo 1 > foo
$ hg add foo
$ hg ci -Am B_0
$ hg log -G
@ changeset: 3:bfcb8e629987
| tag: tip
| parent: 0:06f48e4098b8
| user: test
| date: Thu Jan 01 00:00:00 1970 +0000
| summary: B_0
|
| o changeset: 2:7fd8de258aa4
| | branch: A
| | user: test
| | date: Thu Jan 01 00:00:00 1970 +0000
| | summary: A_1
| |
| o changeset: 1:ae1692b8aadb
|/ branch: A
| user: test
| date: Thu Jan 01 00:00:00 1970 +0000
| summary: A_0
|
o changeset: 0:06f48e4098b8
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: R_0
$ cd ..
Actual tests
============
Initial cloning if needed
-------------------------
$ hg admin::chainsaw-update --dest repo --rev default --source ./src
no such directory: "repo"
creating repository at "repo"
recovering after interrupted transaction, if any
no interrupted transaction available
pulling from ./src
updating to revision 'default'
2 files updated, 0 files merged, 0 files removed, 0 files unresolved
chainsaw-update to revision 'default' for repository at '$TESTTMP/repo' done
$ cd repo
$ hg log -G
@ changeset: 3:bfcb8e629987
| tag: tip
| parent: 0:06f48e4098b8
| user: test
| date: Thu Jan 01 00:00:00 1970 +0000
| summary: B_0
|
| o changeset: 2:7fd8de258aa4
| | branch: A
| | user: test
| | date: Thu Jan 01 00:00:00 1970 +0000
| | summary: A_1
| |
| o changeset: 1:ae1692b8aadb
|/ branch: A
| user: test
| date: Thu Jan 01 00:00:00 1970 +0000
| summary: A_0
|
o changeset: 0:06f48e4098b8
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: R_0
$ hg status -A
C foo
C root
$ cat foo
1
Test lock breacking capabilities
--------------------------------
Demonstrate lock-breaking capabilities with locks that regular Mercurial
operation would not break, because the hostnames registered in locks differ
from the current hostname (happens a lot with succesive containers):
$ ln -s invalid.host.test/effffffc:171814 .hg/store/lock
$ ln -s invalid.host.test/effffffc:171814 .hg/wlock
$ hg debuglock
lock: (.*?), process 171814, host invalid.host.test/effffffc \((\d+)s\) (re)
wlock: (.*?), process 171814, host invalid.host.test/effffffc \((\d+)s\) (re)
[2]
$ hg admin::chainsaw-update --no-purge-ignored --dest . --rev default --source ../src
loaded repository at "."
had to break store lock
had to break working copy lock
recovering after interrupted transaction, if any
no interrupted transaction available
pulling from ../src
updating to revision 'default'
0 files updated, 0 files merged, 0 files removed, 0 files unresolved
chainsaw-update to revision 'default' for repository at '$TESTTMP/repo' done
Test file purging capabilities
------------------------------
Let's also add local modifications (tracked and untracked) to demonstrate the
purging.
$ echo untracked > bar
$ echo modified > foo
$ hg status -A
M foo
? bar
C root
$ echo 2 > ../src/foo
$ hg -R ../src commit -mB_1
$ hg admin::chainsaw-update --dest . --rev default --source ../src -q
no interrupted transaction available
$ hg log -G
@ changeset: 4:973ab81c95fb
| tag: tip
| user: test
| date: Thu Jan 01 00:00:00 1970 +0000
| summary: B_1
|
o changeset: 3:bfcb8e629987
| parent: 0:06f48e4098b8
| user: test
| date: Thu Jan 01 00:00:00 1970 +0000
| summary: B_0
|
| o changeset: 2:7fd8de258aa4
| | branch: A
| | user: test
| | date: Thu Jan 01 00:00:00 1970 +0000
| | summary: A_1
| |
| o changeset: 1:ae1692b8aadb
|/ branch: A
| user: test
| date: Thu Jan 01 00:00:00 1970 +0000
| summary: A_0
|
o changeset: 0:06f48e4098b8
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: R_0
$ hg status -A
C foo
C root
$ cat foo
2
Now behaviour with respect to ignored files: they are not purged if
the --no-purge-ignored flag is passed, but they are purged by default
$ echo bar > .hgignore
$ hg ci -Aqm hgignore
$ echo ignored > bar
$ hg status --all
I bar
C .hgignore
C foo
C root
$ hg admin::chainsaw-update --no-purge-ignored --dest . --rev default --source ../src -q
no interrupted transaction available
$ hg status --all
I bar
C .hgignore
C foo
C root
$ cat bar
ignored
$ hg admin::chainsaw-update --dest . --rev default --source ../src -q
no interrupted transaction available
$ hg status --all
C .hgignore
C foo
C root
$ test -f bar
[1]
test --minimal-initial-cloning variant
--------------------------------------
With `--minimal-initial-cloning`, there is no "requesting all changes"
message. Hence clone bundles would be bypassed (TODO test both cases
# with an actual clone-bundle)
$ cd ..
$ hg admin::chainsaw-update --dest repo2 --rev default --source src --initial-clone-minimal
no such directory: "repo2"
creating repository at "repo2"
recovering after interrupted transaction, if any
no interrupted transaction available
pulling from src
updating to revision 'default'
2 files updated, 0 files merged, 0 files removed, 0 files unresolved
chainsaw-update to revision 'default' for repository at '$TESTTMP/repo2' done
$ cd repo2
$ hg log -G
@ changeset: 2:973ab81c95fb
| tag: tip
| user: test
| date: Thu Jan 01 00:00:00 1970 +0000
| summary: B_1
|
o changeset: 1:bfcb8e629987
| user: test
| date: Thu Jan 01 00:00:00 1970 +0000
| summary: B_0
|
o changeset: 0:06f48e4098b8
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: R_0
$ hg status -A
C foo
C root
$ cat foo
2