Martin von Zweigbergk <martinvonz@google.com> [Fri, 17 May 2019 00:57:57 -0700] rev 42339
convert: don't include file in "files" list if it's added in p2
If the file is from p2, we should clearly compare the flags to what
they were in p2.
Also note that manifest.flags('non-existent') unfortunately returns ''
instead of erroring out.
Differential Revision: https://phab.mercurial-scm.org/D6409
Martin von Zweigbergk <martinvonz@google.com> [Fri, 17 May 2019 11:32:48 -0700] rev 42338
convert: demonstrate broken {files} list in merge commits with file flags
When there is a merge in which the flags for a file from p2 is
non-empty, `hg convert` will incorrectly include that in the
changeset's files list.
Differential Revision: https://phab.mercurial-scm.org/D6408
Matt Harbison <matt_harbison@yahoo.com> [Sat, 18 May 2019 19:56:06 -0400] rev 42337
templater: drop support for old style keywords (API)
These changes originated from several commits over a period of time, so I'm
slightly unsure if this is correct. But the tests pass.
Matt Harbison <matt_harbison@yahoo.com> [Sat, 18 May 2019 19:38:47 -0400] rev 42336
commands: drop support for legacy ^cmd registration (API)
Matt Harbison <matt_harbison@yahoo.com> [Sat, 18 May 2019 19:33:48 -0400] rev 42335
extensions: drop support for extsetup() without `ui` argument (API)
Martin von Zweigbergk <martinvonz@google.com> [Fri, 17 May 2019 11:11:40 -0700] rev 42334
relnotes: mention removed support for mixed log graph lines
This adds release notes for
264a2cbb25d0 (graphmod: remove support for
graph lines mixing parent/grandparent styles (BC), 2018-10-16).
Differential Revision: https://phab.mercurial-scm.org/D6407
Augie Fackler <augie@google.com> [Fri, 17 May 2019 11:03:47 -0400] rev 42333
tests: fix test-clonebundles on recent openbsd
I guess openbsd feels like it needs to stringify this errno in
lowercase and omit the "host" part of "hostname. Okay.
Reported in a big test diff talking about libressl, see 6122. I'm not
flagging this because most of that issue is about a libressl string
change, so this doesn't really make a big difference there.
Differential Revision: https://phab.mercurial-scm.org/D6399
Georges Racinet <georges.racinet@octobus.net> [Thu, 16 May 2019 21:17:14 +0200] rev 42332
rust-python3: compatibility fix for integer conversion
On python3, `to_py_object()` on the usize gives us a PyLong,
whereas it is the generic `PyObject` already on python2, which fits
the `py.None()` default value.
Upcasting to `PyObject` explicitely in all cases solves the issue.
Differential Revision: https://phab.mercurial-scm.org/D6396
Augie Fackler <augie@google.com> [Fri, 17 May 2019 09:42:02 -0400] rev 42331
rust: sort dependencies entries in Cargo.toml
I should probably write a test to enforce this...
Differential Revision: https://phab.mercurial-scm.org/D6398
Pulkit Goyal <7895pulkit@gmail.com> [Fri, 17 May 2019 00:04:29 +0530] rev 42330
py3: make contrib/testparseutil.py to work on str(unicodes)
contrib/check-code work on unicodes and call functions from testparseutil.py
which before this patch used to work on bytes.
This path removes that inconsistency and make testparseutil.py work on unicodes.
This makes test-check-code.t and test-contrib-check-code.t work on Python 3
again.
Differential Revision: https://phab.mercurial-scm.org/D6391
Raphaël Gomès <rgomes@octobus.net> [Fri, 17 May 2019 09:36:29 -0400] rev 42329
rust-filepatterns: call new Rust implementations from Python
This change adds the import to the `rust-cpython` bindings and uses
them when appropriate.
A wrapper function has been defined in the case of `_regex` to
keep this patch simple.
Differential Revision: https://phab.mercurial-scm.org/D6273
Raphaël Gomès <rgomes@octobus.net> [Fri, 17 May 2019 09:36:29 -0400] rev 42328
rust-filepatterns: add `rust-cpython` bindings for `filepatterns`
This change adds the `rust-cpython` interface for top-level functions and
exceptions in the filepatterns module.
Contrary to the Python implementation, this tries to have finer-grained
exceptions to allow for better readability and flow control down the line.
Differential Revision: https://phab.mercurial-scm.org/D6272
Raphaël Gomès <rgomes@octobus.net> [Wed, 24 Apr 2019 11:34:09 +0200] rev 42327
rust-filepatterns: add a Rust implementation of pattern-related utils
This change introduces Rust implementations of two functions related to
pattern handling, all located in `match.py`:
- `_regex`
- `readpatternfile`
These utils are useful in the long-term effort to improve `hg status`'s
performance using Rust. Experimental work done by Valentin Gatien-Baron
shows very promising improvements, but is too different from the current
Mercurial core code structure to be used "as-is".
This is the first - albeit very small - step towards the code revamp
needed down the line.
Two dependencies were added: `regex` and `lazy_static`. Both of them
will be useful for a majority of the Rust code that will be written,
are well known and maintained either by the Rust core team, or by
very frequent contributors.
Differential Revision: https://phab.mercurial-scm.org/D6271
Martin von Zweigbergk <martinvonz@google.com> [Wed, 15 May 2019 22:11:41 -0700] rev 42326
exchange: don't take wlock if bookmarks are stored in .hg/store/
If bookmarks are stored in .hg/store/, there is no need for the
wlock().
Differential Revision: https://phab.mercurial-scm.org/D6388
Martin von Zweigbergk <martinvonz@google.com> [Wed, 15 May 2019 22:09:02 -0700] rev 42325
bookmarks: keep bookmarks in .hg/store if new config set
Bookmarks storage consists of two parts: (1) the set of bookmarks and
their positions, and (2) the current bookmark. The former can get
updated by exchange, while the latter cannot. However, they are both
stored in directly .hg/ and protected by repo.wlock(). As a result,
ugly workarounds were needed. This patch introduces a new config
option to store the set of bookmarks and their positions in .hg/store/
but still storing the current bookmark directory in .hg/. The config
option only takes effect at repo creation time. It results in a new
requirement being set.
Differential Revision: https://phab.mercurial-scm.org/D6387
Pierre-Yves David <pierre-yves.david@octobus.net> [Mon, 20 May 2019 10:08:28 +0200] rev 42324
bookmark: also make bookmark cache depends of the changelog
Since the changelog is also used during the parsing of bookmark data, it should
be listed as a file cache dependency. This fix the race condition we just
introduced a test for.
This is a simple fix that might lead bookmark data to be invalidated more often
than necessary. We could have more complicated code to deal with this race in a
more "optimal" way. I feel it would be unsuitable for stable.
In addition, the performance impact of this is probably minimal and I don't
foresee the more advanced fix to actually be necessary.
Pierre-Yves David <pierre-yves.david@octobus.net> [Mon, 20 May 2019 10:08:17 +0200] rev 42323
localrepo: grab mixedrepostorecache class from
526750cdd02d
On default, Martin von Zweigbergk <martinvonz@google.com> introduced a more
advance filecache decorator. I need this decorator to fix a bug on stable. So I
am grafting the relevant part of
526750cdd02d.
Pierre-Yves David <pierre-yves.david@octobus.net> [Mon, 20 May 2019 10:06:53 +0200] rev 42322
bookmark: add a test for a race condition on push
Bookmark pointing to unknown nodes are ignored. Later these ignored bookmarks
are dropped when writing the file back on disk. On paper, this behavior should
be fine, but with the current implementation, it can lead to unexpected
bookmark deletions.
In theory, to make sure writer as a consistent view, taking the lock also
invalidate bookmark data we already loaded into memory. However this
invalidation is incomplete. The data are stored in a `filecache` that preserve
them if the bookmark related file are untouched. In practice, the bookmark data
in memory also depends of the changelog content, because of the step checking
if the bookmarks refers to a node known to the changelog. So if the bookmark
data were loaded from an up to date bookmark file but filtered with an outdated
changelog file this go undetected.
This condition is fairly specific, but can occurs very often in practice. We
introduce a test recreating the situation. The test comes in an independant
changeset to show it actually reproduce the situation. The fix will come soon
after.
A large share of the initial investigation of this race condition was made by
Valentin Gatien-Baron <valentin.gatienbaron@gmail.com>.
Pierre-Yves David <pierre-yves.david@octobus.net> [Mon, 20 May 2019 07:11:16 +0200] rev 42321
test: properly gate a zstd section
This part of the test can't run if zstd is not available. This was caught by
--pure test (who don't support zstd).
Pierre-Yves David <pierre-yves.david@octobus.net> [Mon, 20 May 2019 07:11:06 +0200] rev 42320
test: update test for expected test output
In
1fac9b931d46 as new test session was introduced. It did not take in account
some part that only ran for pure.
The test is now fixed.
Yuya Nishihara <yuya@tcha.org> [Thu, 16 May 2019 08:15:20 +0900] rev 42319
log: flag topo-sorted set as such
This isn't required right now, but revs.istopo() should return True.
Martin von Zweigbergk <martinvonz@google.com> [Wed, 09 Jan 2019 15:54:45 -0800] rev 42318
copies: fix duplicatecopies() with overlay context
The reasoning for this check is in
78d760aa3607 (duplicatecopies: do
not mark items not in the dirstate as copies, 2013-03-28). The check
was then moved to workingfilectx in
754b5117622f (context: add
workingfilectx.markcopied, 2017-10-15) and no corresponding check was
added later when overlayworkingfilectx was added. Rather than adding
the check there, this patch adds a more generic check on the callers
side and removes the check in workingfilectx.markcopied().
Differential Revision: https://phab.mercurial-scm.org/D6380
Martin von Zweigbergk <martinvonz@google.com> [Wed, 15 May 2019 16:10:52 -0700] rev 42317
tests: demonstrate crash when rebasing across copy with --collapse
As reported by timeless.
Differential Revision: https://phab.mercurial-scm.org/D6379
Augie Fackler <augie@google.com> [Wed, 15 May 2019 17:18:57 -0400] rev 42316
exthelper: add some semi-useful trace logs
It'd be nice to make the trace functions a little better-named in the output,
but I'm not sure how much better we can do without overhead. This at least
lets you see if a single reposetup function is eating all the time or if it's
spread over all of them. I needed this because Google's uber-extension has a
long load time and I wasn't sure where the problem was.
Differential Revision: https://phab.mercurial-scm.org/D6381
Martin von Zweigbergk <martinvonz@google.com> [Wed, 15 May 2019 23:26:05 -0700] rev 42315
help: add missing blank line, making "revlog-compression" show up
Differential Revision: https://phab.mercurial-scm.org/D6386
Martin von Zweigbergk <martinvonz@google.com> [Wed, 15 May 2019 11:53:22 -0700] rev 42314
tests: fix share test to actually share the repo
"repo2" is clearly meant to be a share from "repo1" but without
sharing bookmarks. However, `hg unshare` was called in the repo, so it
had become completely unrelated and thus not testing what it was
supposed to test.
Differential Revision: https://phab.mercurial-scm.org/D6385
Martin von Zweigbergk <martinvonz@google.com> [Wed, 15 May 2019 11:38:45 -0700] rev 42313
tests: separate out bookmarks tests from test-share.t
Differential Revision: https://phab.mercurial-scm.org/D6384
Martin von Zweigbergk <martinvonz@google.com> [Wed, 15 May 2019 10:19:36 -0700] rev 42312
bookmarks: use vfs.tryread() instead of reimplementing it
Differential Revision: https://phab.mercurial-scm.org/D6383
Martin von Zweigbergk <martinvonz@google.com> [Wed, 15 May 2019 10:13:29 -0700] rev 42311
bookmarks: use context manager when writing files
Differential Revision: https://phab.mercurial-scm.org/D6382
timeless <timeless@mozdev.org> [Wed, 15 May 2019 10:54:36 -0400] rev 42310
bisect: do not crash with rewritten commits
Martin von Zweigbergk <martinvonz@google.com> [Wed, 01 May 2019 09:34:47 -0700] rev 42309
log: add config for making `hg log -G` always topo-sorted
I (and everyone else at Google) have an log alias that adds graph mode
and templating. I have another one that builds on the first and also
restricts the set of revisions to only show those I'm most likely to
care about. This second alias also adds topological sorting. I still
sometimes use the first one. When I do, it very often bothers me that
it's not topologically sorted (branches are interleaved). This patch
adds a config option for always using topological sorting with graph
log.
The revision set is sorted eagerly, which seems like a bad idea, but
it doesn't seem to make a big difference in the hg repo (150ms). I
initially tried to instead wrap the user's revset in sort(...,topo),
but that seemed much harder.
Differential Revision: https://phab.mercurial-scm.org/D6331
Martin von Zweigbergk <martinvonz@google.com> [Tue, 14 May 2019 09:13:39 -0700] rev 42308
log: remove an unnecessary "and opts.get('rev')" condition
As Yuya pointed out, the condition is unnecessary since
revs.isdescending() would be true if --follow without --rev.
Differential Revision: https://phab.mercurial-scm.org/D6372
Kyle Lippincott <spectral@google.com> [Tue, 16 Oct 2018 04:59:36 -0700] rev 42307
graphmod: remove support for graph lines mixing parent/grandparent styles (BC)
Currently, if the configuration for a graph edge draw style has multiple bytes
(at least on python2), it is interpreted as "this is a request to draw the line
partially in the style of the parent, partially in the style of the
grandparent". This precludes the configuration handling unicode characters
(which trigger the `len > 1` check, at least on python2), and I believe was part
of the reason that beautifygraph was written the way it was.
Talking with the person who implemented this, it appears to have been to achieve
feature parity with the rendering of the smartlog extension. I suspect that this
isn't actually used outside of that situation, so I think that we can remove it
without much issue.
This will make it so that multi-character edges are possible, and render any
existing configuration that uses this feature with these multiple characters.
This is *not* going to adjust the width of everything to make it line up
correctly, please see the test that's being modified in this changeset for an
example of how the previous configuration now renders.
Note also that the previous configuration seems to have been broken, or at least
it was behaving in a really non-obvious way - it was avoiding the grandparent
character(s) when it should have been displaying them! This is why so many "!"
characters changed to "3."; I don't know if this was intentional.
Differential Revision: https://phab.mercurial-scm.org/D5112
Pulkit Goyal <7895pulkit@gmail.com> [Wed, 15 May 2019 21:02:32 +0300] rev 42306
py3: add 5 new passing tests
Differential Revision: https://phab.mercurial-scm.org/D6378
Pulkit Goyal <7895pulkit@gmail.com> [Wed, 15 May 2019 20:37:39 +0300] rev 42305
py3: add a r'' to prevent transformer adding b''
# skip-blame because just r'' prefix
Differential Revision: https://phab.mercurial-scm.org/D6377
Raphaël Gomès <rgomes@octobus.net> [Mon, 06 May 2019 22:51:10 +0200] rev 42304
rust-dirstate: call parse/pack bindings from Python
A future patch will need to address the issue of Rust module policy,
to avoid having ugly duplicate imports and conditionals all over the place.
As the rewrite of dirstate in Rust progresses, we will need fewer of those
"contact points".
Differential Revision: https://phab.mercurial-scm.org/D6350
Raphaël Gomès <rgomes@octobus.net> [Mon, 06 May 2019 22:50:34 +0200] rev 42303
rust-dirstate: add rust-cpython bindings to the new parse/pack functions
This allows for Python code to call `parse/pack_dirstate` transparently.
These bindings are heavy given the relatively simple task, as they are bound
to implementation details of both the C and Python code. They will be slimmed
down in future patches and eventually completely removed once more of the
dirstate code has been refactored/rewritten in Rust.
Both functions emulate the mutate-on-loop style of the Python and C
implementations by looping over changed items in the compatibility layer,
instead of at the core functions.
Differential Revision: https://phab.mercurial-scm.org/D6349
Raphaël Gomès <rgomes@octobus.net> [Mon, 06 May 2019 22:48:09 +0200] rev 42302
rust-dirstate: add rust implementation of `parse_dirstate` and `pack_dirstate`
Working towards the goal of having a complete Rust implementation of
`hg status`, these two utils are a first step of many to be taken
to improve performance and code maintainability.
Two dependencies have been added: `memchr` and `byteorder`.
Both of them have been written by reputable community members and are
very mature crates.
The Rust code will often need to use their byte-oriented functions.
A few unit tests have been added and may help future development and debugging.
In a future patch that uses `parse_dirstate` to stat the working tree in
parallel - which neither the Python nor the C implementations do - actual
performance improvements will be seen for larger repositories.
Differential Revision: https://phab.mercurial-scm.org/D6348
Martin von Zweigbergk <martinvonz@google.com> [Tue, 14 May 2019 22:56:58 -0700] rev 42301
changelog: define changelogrevision.p[12]copies for null revision
Looks like I missed these in
5382d8f8530b (changelog: parse copy
metadata if available in extras, 2017-12-27). `hg debugp[12]copies -r
null` fails before this patch.
Differential Revision: https://phab.mercurial-scm.org/D6376
Martin von Zweigbergk <martinvonz@google.com> [Tue, 23 Apr 2019 13:29:13 -0700] rev 42300
copies: write empty entries in changeset when also writing to filelog
When writing to both changeset and filelog (during transition), we
don't want the reader to waste time by falling back to reading from
the filelog when there is no copy metadata. Let's write out empty copy
metadata instead (the read path is already prepared for this
case). Thanks to Greg for pointing this out.
Differential Revision: https://phab.mercurial-scm.org/D6306
timeless <timeless@mozdev.org> [Mon, 13 May 2019 14:19:36 -0400] rev 42299
rebase: hide help for revisions.Predicates._destautoorphanrebase
timeless <timeless@mozdev.org> [Fri, 03 May 2019 16:07:57 -0400] rev 42298
unshelve: add space to help
Martin von Zweigbergk <martinvonz@google.com> [Fri, 10 May 2019 22:24:47 -0700] rev 42297
context: default to using branch from dirstate only in workingctx
Same reasoning as previous commits: only the workingctx should know
about the dirstate.
committablectx now seems free of dirstate references.
Differential Revision: https://phab.mercurial-scm.org/D6374
Martin von Zweigbergk <martinvonz@google.com> [Fri, 10 May 2019 22:51:33 -0700] rev 42296
context: let caller pass in branch to committablectx.__init__()
committablectx.__init__() currently looks up the branch from the
dirstate unless it's passed in the extras. memctx.__init__() has a
branch argument, but since committablectx.__init__() doesn't accept
it, it lets that constructor look up the branch from the dirstate
before it overwrites it, which seems awkward.
Differential Revision: https://phab.mercurial-scm.org/D6366
Martin von Zweigbergk <martinvonz@google.com> [Fri, 10 May 2019 21:55:59 -0700] rev 42295
context: move contents of committablectx.markcommitted() to workingctx
Same reasoning as previous commits: this function updates the
dirstate. By not updating the dirstate here, we also fix the
close-head test.
Differential Revision: https://phab.mercurial-scm.org/D6365
Martin von Zweigbergk <martinvonz@google.com> [Fri, 10 May 2019 22:18:11 -0700] rev 42294
tests: demonstrate that close-head command updates working copy
The help text for the command says "...it doesn't change the working
directory", so I don't think this is intentional.
Differential Revision: https://phab.mercurial-scm.org/D6364
Martin von Zweigbergk <martinvonz@google.com> [Fri, 10 May 2019 21:53:41 -0700] rev 42293
context: move walk() and match() overrides from committablectx to workingctx
Same reasoning as previous commit: these functions update the dirstate.
Differential Revision: https://phab.mercurial-scm.org/D6363
Martin von Zweigbergk <martinvonz@google.com> [Fri, 10 May 2019 21:35:30 -0700] rev 42292
context: move flags overrides from committablectx to workingctx
These read from the dirstate, so they shouldn't be used in other
subclasses.
Differential Revision: https://phab.mercurial-scm.org/D6362
Martin von Zweigbergk <martinvonz@google.com> [Fri, 10 May 2019 13:41:42 -0700] rev 42291
context: reuse changectx._copies() in all but workingctx
This moves the dirstate-specific _copies() implementation from
committablectx into workingctx where it should be (I think all
dirstate-specific stuff should be moved into workingctx). The part of
changectx._copies() that is for producing changeset-wide copy dicts
from the filectxs is moved into basectx so it's reused by the other
subclasses. The part of changectx._copies() that's about reading copy
information from the changeset remains there. This fixes in-memory
rebase (and makes `hg convert` able to write copies to changesets).
Differential Revision: https://phab.mercurial-scm.org/D6219
Martin von Zweigbergk <martinvonz@google.com> [Fri, 10 May 2019 14:27:22 -0700] rev 42290
overlayworkingctx: don't include added-then-deleted files in memctx
If a file (such as a .orig file) is temporarily added to the
overlayworkingctx and then deleted, it's still going to be in the
_cache dict. In tomemctx(), we created the list of files from
_cache.keys(), so the memctx.files() would include the temporary
file. That was fine because the list of files was only used in
localrepo.commitctx() (I think), where there's an extra filtering of
incorrectly removed files (annotated with an inaccurate "update
manifest" comment). I'd like to call memctx.files() in another case,
but first we need to make it accurate.
Differential Revision: https://phab.mercurial-scm.org/D6361
Martin von Zweigbergk <martinvonz@google.com> [Fri, 10 May 2019 10:23:46 -0700] rev 42289
tests: demonstrate loss of changeset copy metadata on rebase
Differential Revision: https://phab.mercurial-scm.org/D6360
Martin von Zweigbergk <martinvonz@google.com> [Fri, 10 May 2019 11:03:54 -0700] rev 42288
overlaycontext: allow calling copydata() on clean context
We should just report no copy if the context is clean.
Differential Revision: https://phab.mercurial-scm.org/D6358
Martin von Zweigbergk <martinvonz@google.com> [Fri, 10 May 2019 10:23:08 -0700] rev 42287
tests: demonstrate another failure with in-memory rebase and copies
This is a similar to
dd1ab72be983 (test: demonstrate crash with
in-memory rebase and copies, 2019-03-14). The new failure started with
57203e0210f8 (copies: calculate mergecopies() based on pathcopies(),
2019-04-11). It happens in the call to mergemod.update() on
rebase.py:1268 where we call mergemod.update() to graft a node. Since
the mergecopies() rewrite, that calls _related() with the filectx from
the overlaywctx instead of a filectx from the changectx where the file
was last modified. Either should be fine, so I don't think that's
a bug.
Differential Revision: https://phab.mercurial-scm.org/D6357
Martin von Zweigbergk <martinvonz@google.com> [Tue, 14 May 2019 16:40:49 -0700] rev 42286
commit: fix a typo ("form p1" -> "from p1")
Differential Revision: https://phab.mercurial-scm.org/D6375
Gregory Szorc <gregory.szorc@gmail.com> [Sat, 27 Apr 2019 11:48:26 -0700] rev 42285
automation: initial support for running Linux tests
Building on top of our Windows automation support, this commit
implements support for performing automated tasks on remote Linux
machines. Specifically, we implement support for running tests
on ephemeral EC2 instances. This seems to be a worthwhile place
to start, as building packages on Linux is more or less a solved
problem because we already have facilities for building in Docker
containers, which provide "good enough" reproducibility guarantees.
The new `run-tests-linux` command works similarly to
`run-tests-windows`: it ensures an AMI with hg dependencies is
available, provisions a temporary EC2 instance with this AMI, pushes
local changes to that instance via SSH, then invokes `run-tests.py`.
Using this new command, I am able to run the entire test harness
substantially faster then I am on my local machine courtesy of
access to massive core EC2 instances:
wall: 16:20 ./run-tests.py -l (i7-6700K)
wall: 14:00 automation.py run-tests-linux --ec2-instance c5.2xlarge
wall: 8:30 automation.py run-tests-linux --ec2-instance m5.4xlarge
wall: 8:04 automation.py run-tests-linux --ec2-instance c5.4xlarge
wall: 4:30 automation.py run-tests-linux --ec2-instance c5.9xlarge
wall: 3:57 automation.py run-tests-linux --ec2-instance m5.12xlarge
wall: 3:05 automation.py run-tests-linux --ec2-instance m5.24xlarge
wall: 3:02 automation.py run-tests-linux --ec2-instance c5.18xlarge
~3 minute wall time to run pretty much the entire test harness is
not too bad!
The AMIs install multiple versions of Python. And the run-tests-linux
command specifies which one to use:
automation.py run-tests-linux --python system3
automation.py run-tests-linux --python 3.5
automation.py run-tests-linux --python pypy2.7
By default, the system Python 2.7 is used. Using this functionality,
I was able to identity some unexpected test failures on PyPy!
Included in the feature is support for running with alternate
filesystems. You can simply pass --filesystem to the command to
specify the type of filesystem to run tests on. When the ephemeral
instance is started, a new filesystem will be created and tests
will run from it:
wall: 4:30 automation.py run-tests-linux --ec2-instance c5.9xlarge
wall: 4:20 automation.py run-tests-linux --ec2-instance c5d.9xlarge --filesystem xfs
wall: 4:24 automation.py run-tests-linux --ec2-instance c5d.9xlarge --filesystem tmpfs
wall: 4:26 automation.py run-tests-linux --ec2-instance c5d.9xlarge --filesystem ext4
We also support multiple Linux distributions:
$ automation.py run-tests-linux --distro debian9
total time: 298.1s; setup: 60.7s; tests: 237.5s; setup overhead: 20.4%
$ automation.py run-tests-linux --distro ubuntu18.04
total time: 286.1s; setup: 61.3s; tests: 224.7s; setup overhead: 21.4%
$ automation.py run-tests-linux --distro ubuntu18.10
total time: 278.5s; setup: 58.2s; tests: 220.3s; setup overhead: 20.9%
$ automation.py run-tests-linux --distro ubuntu19.04
total time: 265.8s; setup: 42.5s; tests: 223.3s; setup overhead: 16.0%
Debian and Ubuntu are supported because those are what I use and am
most familiar with. It should be easy enough to add support for other
distros.
Unlike the Windows AMIs, Linux EC2 instances bill per second. So
the cost to instantiating an ephemeral instance isn't as severe.
That being said, there is some overhead, as it takes several dozen
seconds for the instance to boot, push local changes, and build
Mercurial. During this time, the instance is largely CPU idle and
wasting money. Even with this inefficiency, running tests is
relatively cheap: $0.15-$0.25 per full test run. A machine running
tests as efficiently as these EC2 instances would cost say $6,000, so
you can run the test harness a >20,000 times for the cost of an
equivalent machine. Running tests in EC2 is almost certainly cheaper
than buying a beefy machine for developers to use :)
# no-check-commit because foo_bar function names
Differential Revision: https://phab.mercurial-scm.org/D6319
Gregory Szorc <gregory.szorc@gmail.com> [Tue, 23 Apr 2019 21:57:32 -0700] rev 42284
automation: move image operations to own functions
An upcoming commit will need this functionality with slightly different
values and it is enough code to not want to duplicate. Let's refactor
into standalone functions so it can be reused.
Differential Revision: https://phab.mercurial-scm.org/D6318
Gregory Szorc <gregory.szorc@gmail.com> [Fri, 19 Apr 2019 09:18:23 -0700] rev 42283
automation: add --version argument to build-all-windows-packages
This lets us pass a version string through when building all
Windows packages, just like we can do with the individual commands
which produce installers.
Differential Revision: https://phab.mercurial-scm.org/D6317
Gregory Szorc <gregory.szorc@gmail.com> [Fri, 19 Apr 2019 08:32:24 -0700] rev 42282
automation: do a force push to synchronize
We don't know what the state of the remote is. Force pushing will
be more resilient.
Differential Revision: https://phab.mercurial-scm.org/D6316
Gregory Szorc <gregory.szorc@gmail.com> [Fri, 19 Apr 2019 08:21:02 -0700] rev 42281
automation: add check that hg source directory is a repo
Synchronizing from e.g. source distributions is not yet supported.
Let's add a check so we fail with an error message indicating
such.
Differential Revision: https://phab.mercurial-scm.org/D6315
Gregory Szorc <gregory.szorc@gmail.com> [Fri, 19 Apr 2019 07:34:55 -0700] rev 42280
automation: shore up rebooting behavior
There was a race condition in the old code. Use
instance.stop()/instance.start() to eliminate it.
As part of debugging this, I also found another race condition
related to PowerShell permissions after the reboot. Unfortunately,
I'm not sure the best way to work around it. I've added a comment
for now.
Differential Revision: https://phab.mercurial-scm.org/D6288
Gregory Szorc <gregory.szorc@gmail.com> [Fri, 19 Apr 2019 06:07:00 -0700] rev 42279
automation: wait longer for WinRM connection
I got a few timeouts waiting for only 120s for the WinRM connection
to become available. Increasing to 180s seems to fix. I guess
AWS isn't as consistent as I would like :(
Differential Revision: https://phab.mercurial-scm.org/D6287
Gregory Szorc <gregory.szorc@gmail.com> [Sat, 27 Apr 2019 11:38:58 -0700] rev 42278
automation: wait for instance profiles and roles
Otherwise there is a race condition between creating the resources
and us attempting to use them / them becoming available.
The role waiter API was recently introduced, so we had to upgrade
the boto3 package to get it. Other packages were also updated
to latest versions just because.
Even with this change, I still run into issues with the IAM instance
profile not being available when we attempt to create an EC2 instance
using a just-created profile. I'm not sure what's going on. Possibly
a bug on Amazon's end. But the new behavior is "more correct."
Differential Revision: https://phab.mercurial-scm.org/D6286
Gregory Szorc <gregory.szorc@gmail.com> [Fri, 19 Apr 2019 05:20:33 -0700] rev 42277
automation: don't create resources when deleting things
Otherwise running these commands can result in resources being
created. In the case of `purge-ec2-resources`, we will create
resources only to delete them immediately afterwards!
With this change, `purge-ec2-resources` now no-ops if no
resources exist.
# no-check-commit because foo_bar function name
Differential Revision: https://phab.mercurial-scm.org/D6285
Gregory Szorc <gregory.szorc@gmail.com> [Fri, 19 Apr 2019 05:15:43 -0700] rev 42276
automation: detach policies before deleting role
You can't delete an IAM role that has attached policies.
With this change, the purge-ec2-resources command now works.
Differential Revision: https://phab.mercurial-scm.org/D6284
Gregory Szorc <gregory.szorc@gmail.com> [Fri, 19 Apr 2019 05:07:44 -0700] rev 42275
automation: only iterate over our AMIs
We can't delete AMIs that we don't own. Iterating over other
AMIs won't work and slows down execution.
Differential Revision: https://phab.mercurial-scm.org/D6283
Martin von Zweigbergk <martinvonz@google.com> [Wed, 01 May 2019 15:34:03 -0700] rev 42274
remotefilelog: move most setup from onetimesetup() to uisetup()
All the wrappers moved in this patch check if remotefilelog is enabled
before they change behavior, so it's safe to always wrap.
Differential Revision: https://phab.mercurial-scm.org/D6334
Martin von Zweigbergk <martinvonz@google.com> [Wed, 01 May 2019 15:24:16 -0700] rev 42273
remotefilelog: move most functions in onetimeclientsetup() to top level
This is how most extensions seem to do it. It makes sure we don't
accidentally depend on the captured ui instance.
Differential Revision: https://phab.mercurial-scm.org/D6333
Martin von Zweigbergk <martinvonz@google.com> [Tue, 14 May 2019 09:46:38 -0700] rev 42272
tests: avoid the word "dirty" to mean "not a descendant of merge base"
The term "dirty" is no longer used in the code since
57203e0210f8
(copies: calculate mergecopies() based on pathcopies(), 2019-04-11).
Differential Revision: https://phab.mercurial-scm.org/D6373
Martin von Zweigbergk <martinvonz@google.com> [Wed, 01 May 2019 20:54:27 -0700] rev 42271
releasenotes: add a file in which to record release notes
I've just spent a few very boring hours going through the changelog
for the 5.0 release (829 commits). We only had 5 commits that used the
syntax that the release notes extension expects. This commit adds a
file in which we can record important changes. The file should
preferably be edited in the patch that makes the important change, but
it can also be edited after (I think this is an important benefit
compared to the release notes extension).
I'm thinking that we can rename the file from "next" to "5.1" or
something when it's time, and then we'd create a new "next" file on
the default branch.
I've used the syntax that we use on the our wiki in the template, but
I don't care much that we use any valid syntax at all. The idea is
mostly to record important changes when they happen. I expect that
some copy editing will be needed at release time anyway.
Differential Revision: https://phab.mercurial-scm.org/D6332
Matt Harbison <matt_harbison@yahoo.com> [Sat, 11 May 2019 22:08:57 -0400] rev 42270
record: avoid modifying the matcher passed as a method parameter
No problem observed, but I remember the previous pattern causing problems with
largefiles and/or subrepos. This special matcher was added in
419ac63fe29c, so
directly modifying the `fail` callback was probably an oversight in
44611ad4fbd9.
Differential Revision: https://phab.mercurial-scm.org/D6371
Augie Fackler <augie@google.com> [Sat, 04 May 2019 23:31:42 -0400] rev 42269
sslutil: add support for SSLKEYLOGFILE to wrapsocket
I recently learned of a Firefox/Chrome feature that allows
wiresharking otherwise-TLS'd network connections. Gloriously, there's
a pypi module that enables this same feature on Python, so let's add
support for it to Mercurial in case we need to wireshark some HTTPs
connections.
Differential Revision: https://phab.mercurial-scm.org/D6343
Ian Moody <moz-ian@perix.co.uk> [Sun, 05 May 2019 17:04:48 +0100] rev 42268
phabricator: add custom vcr matcher to match request bodies
Currently when the phabricator extension's conduit output changes the tests
don't notice since the default vcr matcher only matches on 'method' and 'uri',
not the body.
Add a custom matcher that checks the same params are in the body (ignoring
ordering).
vcr's in-built body matcher can't be used since it fails under py3 with a
"UnicodeEncodeError" on the "€ in commit message" tests.
The DREV ids have decreased since the recordings were generated against a
different phabricator instance to avoid spamming mercurial-devel.
Differential Revision: https://phab.mercurial-scm.org/D6347
Augie Fackler <augie@google.com> [Thu, 09 May 2019 18:37:37 -0400] rev 42267
merge with stable
Martin von Zweigbergk <martinvonz@google.com> [Wed, 08 May 2019 21:25:23 -0700] rev 42266
absorb: be more specific when erroring out on merge commit
When you have a merge commit checked out and run `hg absorb`, it would
tell you
abort: no mutable changeset to change
That makes it sound like the problem is public commits when isn't
really. Let's be more specific in this case.
There was already a test case that test this, so that now prints the
new message. I added a new test case that shows the old message (when
a public commit is checked out).
Differential Revision: https://phab.mercurial-scm.org/D6354
Augie Fackler <augie@google.com> [Wed, 08 May 2019 18:11:33 -0400] rev 42265
remotefilelog: log when we're about to fetch files
I'm debugging a slow client situation and knowing how many files are
in the batch request would be a nice thing.
Differential Revision: https://phab.mercurial-scm.org/D6353
Yuya Nishihara <yuya@tcha.org> [Tue, 30 Apr 2019 15:15:57 +0900] rev 42264
revset: populate wdir() by its hash or revision number
It belongs to the same category as the null hash/revision, and we do handle
these virtual identifiers in id()/rev() predicates. Let's do that more
consistently.
Augie Fackler <augie@google.com> [Wed, 08 May 2019 16:09:50 -0400] rev 42263
sslutil: fsencode path returned by certifi (
issue6132)
By inspection, this is the only codepath that could be returning a
string instead of a bytes on Python 3.
Yuya Nishihara <yuya@tcha.org> [Tue, 30 Apr 2019 15:10:07 +0900] rev 42262
revset: extract private constant of {nullrev, wdirrev} set
I'll add a few more users of this constant to get around wdir identifiers.
Yuya Nishihara <yuya@tcha.org> [Tue, 30 Apr 2019 15:22:03 +0900] rev 42261
help: suggest merge() revset instead of -m/--only-merges
Suggested by Dr Rainer Woitok.
Martin von Zweigbergk <martinvonz@google.com> [Mon, 06 May 2019 22:06:23 -0700] rev 42260
tests: update annotate tests to work around simplemerge bug
test-annotate.t and test-fastannotate.hg were failing with --pure
since
57203e0210f8 (copies: calculate mergecopies() based on
pathcopies(), 2019-04-11). It turned out to be because the pure file
merge code behaved differently. I'm guessing it's the
mdiff.get_matching_blocks() that behaves differently, but I haven't
confirmed that.
With this content in the base:
a
a
a
And this on the local side:
a
z
a
And this on the other side:
a
a
a
b4
c
b6
It produced this conflict:
a
z
a
<<<<<<< working copy:
b80e3e32f75a - test: c
||||||| base
a
=======
a
b4
c
b5
>>>>>>> merge rev:
64afcdf8e29e - test: mergeb
I don't care enough about the pure Python code to fix it, so this
patch just updates the tests to manually resolve the conflict.
Differential Revision: https://phab.mercurial-scm.org/D6351
Martin von Zweigbergk <martinvonz@google.com> [Tue, 07 May 2019 14:42:15 -0700] rev 42259
copies: delete misplaced comment
The comment was added in
78d760aa3607 (duplicatecopies: do not mark
items not in the dirstate as copies, 2013-03-28). It became misplaced
in
3666331164bb (cmdutil: add copy-filtering support to
duplicatecopies, 2014-06-07). Then the relevant code was moved far
away in
754b5117622f (context: add workingfilectx.markcopied,
2017-10-15).
Differential Revision: https://phab.mercurial-scm.org/D6352
Ian Moody <moz-ian@perix.co.uk> [Mon, 22 Apr 2019 18:55:27 +0100] rev 42258
phabricator: include branch in the phabread output
Depends on D6301
Differential Revision: https://phab.mercurial-scm.org/D6302
Ian Moody <moz-ian@perix.co.uk> [Mon, 22 Apr 2019 18:55:26 +0100] rev 42257
phabricator: fallback to reading metadata from diff for phabread
Differential Revision: https://phab.mercurial-scm.org/D6301
Ian Moody <moz-ian@perix.co.uk> [Sat, 20 Apr 2019 16:11:23 +0100] rev 42256
phabricator: include commit (node) and parent in the local:commits metadata
Differential Revision: https://phab.mercurial-scm.org/D6298
Martin von Zweigbergk <martinvonz@google.com> [Thu, 18 Apr 2019 00:34:45 -0700] rev 42255
copies: remove redundant filtering of ping-pong renames in _chain()
We already handle the ping-pong rename case in the filtering step, so
there's very little point in doing it in the chaining loop (ping-pong
renames are very rare, so I'm not worried about the cost of adding it
and then removing it again).
Differential Revision: https://phab.mercurial-scm.org/D6344
Augie Fackler <augie@google.com> [Fri, 03 May 2019 15:43:44 -0400] rev 42254
repair: reword comments that I noticed while working on source formatting
I think this is clearer, and one will also keep us from upsetting
check-code when other formatting cleanups happen.
Differential Revision: https://phab.mercurial-scm.org/D6339
Matt Harbison <matt_harbison@yahoo.com> [Mon, 06 May 2019 22:10:34 -0400] rev 42253
commit: allow --interactive to work again when naming a directory (
issue6131)
Sietse Brouwer <sbbrouwer@gmail.com> [Fri, 26 Apr 2019 12:41:48 +0200] rev 42252
gendoc: nest command headers under category headers
Differential Revision: https://phab.mercurial-scm.org/D6329
Sietse Brouwer <sbbrouwer@gmail.com> [Fri, 26 Apr 2019 12:40:26 +0200] rev 42251
minirst: support subsubsubsubsections (header level 5) with marker ''''
Differential Revision: https://phab.mercurial-scm.org/D6328
Sietse Brouwer <sbbrouwer@gmail.com> [Fri, 03 May 2019 15:37:08 +0200] rev 42250
gendoc: guarantee that all commands were processed
The new logic renders the commands belonging to each category in turn.
Commands with an unregistered category are at risk of getting skipped
because their category is not in the list. By comparing the list of all
commands to a log of processed commands, we can detect commands with
unregistered categories and fail with an error message.
Differential Revision: https://phab.mercurial-scm.org/D6327
Sietse Brouwer <sbbrouwer@gmail.com> [Fri, 26 Apr 2019 17:53:01 +0200] rev 42249
gendoc: group commands by category in man page and HTML help
Make Mercurial's man page and HTML help group commands by category, and
present the categories in a helpful order. `hg help` already does this;
this patch uses the same metadata.
This patch uses the same header level for command categories and for
commands. A subsequent patch will push the command headers down one
level.
Differential Revision: https://phab.mercurial-scm.org/D6326
Sietse Brouwer <sbbrouwer@gmail.com> [Thu, 25 Apr 2019 19:15:17 +0200] rev 42248
gendoc: indent loop to make next patch more legible
Differential Revision: https://phab.mercurial-scm.org/D6325
Augie Fackler <augie@google.com> [Fri, 03 May 2019 15:53:56 -0400] rev 42247
contrib: have byteify-strings explode if run in Python 2
Differential Revision: https://phab.mercurial-scm.org/D6341
Augie Fackler <augie@google.com> [Fri, 03 May 2019 15:46:09 -0400] rev 42246
repair: reword comment about bookmarks logic
Again, this will help auto-formatting shortly.
Differential Revision: https://phab.mercurial-scm.org/D6340
Augie Fackler <augie@google.com> [Fri, 03 May 2019 15:42:13 -0400] rev 42245
monotone: fix a bogus _() wrapper that was caught when formatting code
There was a spurious space after `debug`, which hid the _() inside
ui.debug() from check-code. Sigh.
While here, wrap things more concisely.
Differential Revision: https://phab.mercurial-scm.org/D6338
Anton Shestakov <av6@dwimlabs.net> [Fri, 03 May 2019 14:11:16 +0800] rev 42244
commit: add ability to print file status after each successful invocation
When commands.commit.post-status is enabled, `hg commit` will effectively run
`hg status -mardu` after committing. It can help catch mistakes like not
committing all needed files or not adding unknown files that should've been
part of the just created commit.
Anton Shestakov <av6@dwimlabs.net> [Fri, 03 May 2019 14:07:14 +0800] rev 42243
tests: flatten repo structure in test-commit.t
Let's move to parent directory before `hg init` repos, since they don't need to
be nested. It makes amend/strip messages that include full path to the backup
bundle shorter, for instance.
Matt Harbison <matt_harbison@yahoo.com> [Sat, 04 May 2019 01:16:42 -0400] rev 42242
lfs: add a TODO file
This is a cleaned up and reorganized list of items I sent out about a year ago.
But tracking this in the repo (like the narrow extension) gives more visibility
in case anyone wants to help out.
Martin von Zweigbergk <martinvonz@google.com> [Sat, 27 Apr 2019 22:08:45 -0700] rev 42241
copies: make "limit" argument to _tracefile() mandatory
We always pass a limit. I think the fact that it was optional was also
the reason we checked ">=limit" before we used it. So now we can
remove that condition too.
Differential Revision: https://phab.mercurial-scm.org/D6335
Martin von Zweigbergk <martinvonz@google.com> [Fri, 03 May 2019 08:37:10 -0700] rev 42240
localrepo: don't use defaults arguments that will never be overridden
The commithook() callback will be called when the lock is
released. lock.release() calls the callback without arguments, so it
was quite confusing to me that this function declared extra
arguments. We can just close on the variables in the outer scope
instead.
Differential Revision: https://phab.mercurial-scm.org/D6336
Martin von Zweigbergk <martinvonz@google.com> [Fri, 03 May 2019 12:32:00 -0700] rev 42239
tags: avoid double-reversing a list
Differential Revision: https://phab.mercurial-scm.org/D6337
Pierre-Yves David <pierre-yves.david@octobus.net> [Mon, 11 Mar 2019 02:35:18 +0100] rev 42238
updatecaches: also warm hgtagsfnodescache
Now that a full update of this cache run in a reasonable amount of time, we can
warm everything when during a full update.
Pierre-Yves David <pierre-yves.david@octobus.net> [Mon, 11 Mar 2019 01:10:20 +0100] rev 42237
hgtagsfnodescache: inherit fnode from parent when possible
If a changeset does not update the content of `.hgtags`, it means it will use
the same file-node (for `.hgtags`) as its parents. In this case we can
directly reuse the parent's file-node.
We use this property when updating the `hgtagsfnodescache` taking a faster path
if we already have a cached value for the parents of the node we are looking
at.
Doing so provides a large performance boost when looking at a lot of fnodes,
especially on repository with very large manifest:
timing for `tagsmod.fnoderevs(ui, repo, repo.changelog.revs())`
mercurial: (41907 revisions, 1923 files)
before: 6.9 seconds
after: 2.7 seconds (-54%)
pypy: (96266 revisions, 5198 files)
before: 80 seconds
after: 20 seconds (-75%)
mozilla-central: (463411 revisions, 272080 files)
before: 7166.4 seconds
after: 47.8 seconds (-99%, x150 speedup)
On a copy of mozilla-try with about 35K heads ans 1.7M changesets, this moves
the computation from many hours to a couple of minutes, making it more
interesting to do a full warm up of this cache before computing tags (from a
cold cache).
There seems to be other performance low hanging fruits, like avoiding the use of
changectx or a more revision centric logic. However, the new code is fast enough
for my needs right now.
Pierre-Yves David <pierre-yves.david@octobus.net> [Mon, 11 Mar 2019 01:09:38 +0100] rev 42236
hgtagsfnodescache: handle nullid lookup
The null revision is empty, so it `.hgtags` content is `nullid` in regards with
the `hgtagsfnodescache`. Dealing with `nullid` will help with the next
changeset. Before this change, feeding `nullid` to `hgtagsfnodescache.getfnode` would
return a wrong result (fnode for tip).
Sietse Brouwer <sbbrouwer@gmail.com> [Fri, 26 Apr 2019 17:39:07 +0200] rev 42235
help: register the 'gpg' command category and give it a description
help.py expects extensions to register their command category in the
CATEGORY_ORDER and CATEGORY_NAMES variables. Once gendoc.py orders
commands by category, in the next patch, it'll assume this registration
(and raise an exception on encountering any unregistered categories).
Luckily, gpg is the only bundled extension with an unregistered custom
category, so let's fix it.
Differential Revision: https://phab.mercurial-scm.org/D6324
feyu@google.com [Thu, 25 Apr 2019 15:30:40 -0700] rev 42234
histedit: Speed up scrolling in patch view mode
Store patchcontents into the mode state, avoiding the expensive
call to ui for computing the patchcontents.
Before this change in large repos histedit patch view mode can
be very irresponsive.
Yu Feng <rainwoodman@gmail.com> [Thu, 02 May 2019 16:43:34 -0700] rev 42233
histedit: Show file names in multiple line format
Yuya Nishihara <yuya@tcha.org> [Fri, 03 May 2019 20:06:03 +0900] rev 42232
parser: fix crash by parsing "()" in keyword argument position
A tree node can be either None or a tuple because x=("group", None) is
reduced to x[1].
Pierre-Yves David <pierre-yves.david@octobus.net> [Sat, 06 Apr 2019 17:46:19 +0200] rev 42231
repoview: introduce a `experimental.extra-filter-revs` config
The option define revisions to additionally filter out of all repository "view".
The end goal is to provide and easy to way to serve multiple subset of the same
repository using multiple "shares".
The simplest use case of this feature is to have one view serving the public
changesets and one view also serving the draft. This is currently achievable
using the new `server.view` option introduced recently by Joerg Sonnenberger.
However, more advanced use cases need more advanced definitions. For example
some needs a view dedicated to some release branches, or view that hides
security fixes to be released. Joerg Sonnenberger and I discussed this topic at
the recent mini-sprint and the both of us have seen real life use cases for
this. (This series got written during the same mini-sprint).
The feature is fully functional, and use similar cache-fallback mechanism to
ensure decent performance. However,there remaining room to ensure each share
caches and hooks collaborate with each others. This will come at a later time
once users start to actually test this feature on real usecase.
Martin von Zweigbergk <martinvonz@google.com> [Wed, 17 Apr 2019 23:10:29 -0700] rev 42230
copies: filter out copies from non-existent source later in _chain()
_changesetforwardcopies() repeatedly calls _chain(). That is very
expensive because _chain() does lookups in the manifest. I hope to
split up the function in two parts: 1) simple chaining, not
considering end points, and 2) filter out files that don't exist in
the end points (and ping-pong copies/renames).
This patches gets us closer to that by moving the check for
non-existent source later in the function. Now there are no more
checks for "src" and "dst" in the first loop; all the filtering of
invalid copies is done in the second loop. The code also looks much
more consistent now.
No measureable impact on `hg debugpathcopies 4.0 4.8`. That shouldn't
be surprising since the only case we're doing more checks now is in
case of chained copies/renames, which are quire rare in practice.
Differential Revision: https://phab.mercurial-scm.org/D6277
Martin von Zweigbergk <martinvonz@google.com> [Thu, 18 Apr 2019 00:12:56 -0700] rev 42229
copies: clarify mutually exclusive cases in _chain() with a s/if/elif/
If the 'b' dict has a rename from 'x' to 'y', it shouldn't be possible
for 'x' to be both (a key) in 'a' and in 'src'. That would mean that
'x' is a file in the source commit and also a rename destination in
the intermediate commit. But we currently don't allow renaming files
onto existing files, so that shouldn't happen. So let's clarify that
by using an "elif" instead of an "if". And if we did allow renaming
files onto existing files, we should prefer to use the rename
destination in the intermediate commit as source anyway.
Differential Revision: https://phab.mercurial-scm.org/D6276
Martin von Zweigbergk <martinvonz@google.com> [Thu, 18 Apr 2019 00:05:05 -0700] rev 42228
copies: delete a redundant cleanup step in _chain()
The check is redundant since
d5edb5d3a337 (copies: filter out copies
when target is not in destination manifest, 2019-02-14). To test that
hypothesis, I made this change in the commit that commit, but all
tests still passed. I think the case was necessary before then, we
just didn't have tests for it.
Differential Revision: https://phab.mercurial-scm.org/D6275
Martin von Zweigbergk <martinvonz@google.com> [Wed, 17 Apr 2019 23:10:14 -0700] rev 42227
copies: document cases in _chain()
Differential Revision: https://phab.mercurial-scm.org/D6274
Martin von Zweigbergk <martinvonz@google.com> [Wed, 17 Apr 2019 14:44:18 -0700] rev 42226
copies: ignore heuristics copytracing when using changeset-centric algos
Differential Revision: https://phab.mercurial-scm.org/D6269
Martin von Zweigbergk <martinvonz@google.com> [Wed, 17 Apr 2019 14:42:23 -0700] rev 42225
copies: move check for experimental.copytrace==<falsy> earlier
I'm going to ignore experimental.copytrace when changeset-centric
algorithms are required. This little refactoring makes that easier to
add.
Differential Revision: https://phab.mercurial-scm.org/D6268
Martin von Zweigbergk <martinvonz@google.com> [Wed, 17 Apr 2019 14:11:54 -0700] rev 42224
copies: replace .items() by .values() where appropriate
As pointed out by Pierre-Yves.
Differential Revision: https://phab.mercurial-scm.org/D6266
Martin von Zweigbergk <martinvonz@google.com> [Fri, 12 Apr 2019 10:44:37 -0700] rev 42223
copies: inline _computenonoverlap() in mergecopies()
We now call pathcopies() from the base to each of the commits, and
that calls _computeforwardmissing(), which does file prefetching (in
the remotefilelog override). So the call to _computenonoverlap() is
now pointless (the sets of files from _computenonoverlap() are subsets
of the sets of files from _computeforwardmissing()).
This somehow also fixes a broken remotefilelog test.
Differential Revision: https://phab.mercurial-scm.org/D6256
Martin von Zweigbergk <martinvonz@google.com> [Thu, 11 Apr 2019 23:22:54 -0700] rev 42222
copies: calculate mergecopies() based on pathcopies()
When copies are stored in changesets, we need a changeset-centric
version of mergecopies() just like we have a changeset-centric version
of pathcopies(). I think the natural way of thinking about
mergecopies() is in terms of pathcopies() from the base to each of the
commits. So if we can rewrite mergecopies() based on two such
pathcopies() calls, we'll get the changeset-centric version for
free. That's what this patch does.
A nice bonus is that it ends up being a lot simpler. mergecopies() has
accumulated a lot of technical debt over time. One good example is the
code for dealing with grafts (the "partial/incomplete/dirty"
stuff). Since pathcopies() already deals with backwards renames and
ping-pong renames, we get that for free.
I've run tests with hard-coded debug logging for "fullcopy" and while
I haven't looked at every difference it produces, all the ones I have
looked at seemed reasonable to me. I'm a little surprised that no more
tests fail when run with '--extra-config-opt
experimental.copies.read-from=compatibility' compared to before this
patch. This patch also fixes the broken cases in test-annotate.t and
test-fastannotate.t. It also enables the part of test-copies.t that
was previously disabled exactly because mergecopies() needed to get a
changeset-centric version.
One drawback of the rewritten code is that we may now make
remotefilelog prefetch more files. We used to prefetch files that were
unique to either side of the merge compared to the other. We now
prefetch files that are unique to either side of the merge compared to
the base. This means that if you added the same file to each side, we
would not prefetch it before, but we would now. Such cases are
probably quite rare, but one likely scenario where they happen is when
moving from a commit to its successor (or the other way around). The
user will probably already have the files in the cache in such cases,
so it's probably not a big deal.
Some timings for calculating mergecopies between two revisions
(revisions shown on each line, all using the common ancestor as base):
In the hg repo:
4.8 4.9: 0.21s -> 0.21s
4.0 4.8: 0.35s -> 0.63s
In and old copy of the mozilla-unified repo:
FIREFOX_BETA_60_BASE^ FIREFOX_BETA_60_BASE: 0.82s -> 0.82s
FIREFOX_NIGHTLY_59_END FIREFOX_BETA_60_BASE: 2.5s -> 2.6s
FIREFOX_BETA_59_END FIREFOX_BETA_60_BASE: 3.9s -> 4.1s
FIREFOX_AURORA_50_BASE FIREFOX_BETA_60_BASE: 31s -> 33s
So it's measurably slower in most cases. The most significant
difference is in the hg repo between revisions 4.0 and 4.8. In that
case it seems to come from the fact that pathcopies() uses
fctx.isintroducedafter() (in _tracefile), while the old mergecopies()
used fctx.linkrev() (in _checkcopies()). That results in a single call
to filectx._adjustlinkrev(), which is responsible for the entire
difference in time (in my repo). So we pay a performance penalty but
we get more correct code (see change in
test-mv-cp-st-diff.t). Deleting the "== f.filenode()" in _tracefile()
recovers the lost performance in the hg repo.
There were are few other optimizations in _checkcopies() that I could
not measure any impact from. One was from the "seen" set. Another was
from a "continue" when the file was not in the destination manifest
(corresponding to "am" in _tracefile).
Also note that merge copies are not calculated when updating with a
clean working copy, which is probably the most common case. I
therefore think the much simpler code is worth the slowdown.
Differential Revision: https://phab.mercurial-scm.org/D6255
Martin von Zweigbergk <martinvonz@google.com> [Mon, 29 Apr 2019 14:38:54 -0700] rev 42221
tests: add test where copy source is deleted and added back
This shows another difference between pathcopies() and mergecopies():
mergecopies() considers files that have been deleted and then added
back as different files, but pathcopies() does not.
Differential Revision: https://phab.mercurial-scm.org/D6330
Augie Fackler <augie@google.com> [Wed, 01 May 2019 14:30:25 -0400] rev 42220
merge with stable