- Jan 21, 2019
-
-
Valentin Gatien-Baron authored
They are some extra computation that will shortcut the delta compression if the delta seems hopeless, returning None.
-
Valentin Gatien-Baron authored
They are some more optimization change that will make use of this in the function. So we retrieve the data earlier.
-
- Apr 25, 2019
-
-
Valentin Gatien-Baron authored
Before computing any delta, we get a basic estimation of the delta size we can expect and the resulted compressed value. We then checks this projected size against the `size(snapshotⁿ) > size(snapshotⁿ⁺¹)` constraint. This allows to exclude potential base candidates before doing any expensive computation. This only apply to the intermediate-snapshot case since this constraint only apply to them. For some pathological cases of a private repository this step provide a significant performance boost (timing from `hg perfrevlogwrite`): before: 14.115908 seconds after: 3.145906 seconds
-
Valentin Gatien-Baron authored
Before computing any delta, we get a basic estimation of the delta size we can expect and the resulted compressed value. We then checks this projected size against the ½ⁿ size constraints. This allows to exclude potential base candidates before doing any expensive computation. This only apply to the intermediate-snapshot case since this constraint only apply to them. In practice we only perform this new checks for the manifestlog. Manifest log combine two property: it is likely to have delta chain issue and its diffing/compression is fairly predictable. The initial author of this changeset is Valentin Gatien-Baron providing the initial idea and initial testing, Pierre-Yves David later consolidated the code in the right location and run more extensive testing.
-
Pierre-Yves David authored
There are various optimization we can do if we can estimate the size of delta before actually spending CPU compressing them. So we add a attributed dedicated to tracking that. We only use it on Manifest because (1) it structure is quite stable across all Mercurial repository so its compression ratio is fairly universal. This is the revlog with most extreme delta (cf the sparse-revlog optimization). This will be put to use in later changesets. Right now the compression upper bound is set to 10. This is a fairly conservative value (observed value is more around 3), but I prefer to be safe while introducing the optimization principles. We can tune the optimization threshold later.
-
- Jun 12, 2019
-
-
Pierre-Yves David authored
This reduce the chance of developers being surprised by special cases.
-
Pierre-Yves David authored
The help text was copy pasted from the previous option.
-
- May 16, 2019
-
-
Georges Racinet authored
This is more flexible in the passing of additional flags, also what setuptools_rust does, giving less uncertainty about non-Linux platforms.
-
- Jun 14, 2019
-
-
Georges Racinet authored
MacOSX needs special link flags. Quoting the README of rust-cpython: create a `.cargo/config` with the following content: ``` [target.x86_64-apple-darwin] rustflags = [ "-C", "link-arg=-undefined", "-C", "link-arg=dynamic_lookup", ] ``` This is tested with Python 2.7 (Anaconda install) and Python 3 (Homebrew install)
-
Georges Racinet authored
Before this changeset, the shared library objects suffixes were both (rustc output and Python input) hardcoded to '.so', which is wrong for Python3 and non Linux targets.
-
- May 27, 2019
-
-
sliquister authored
The problem is that hg merge/update/etc work the following way: 1. figure out what files to update 2. apply the update to disk 3. apply the update to in-memory dirstate 4. write dirstate where step3 looks at the filesystem and assumes it sees the result of step2. If a file is changed between step2 and step3, step3 will record incorrect information in the dirstate. I avoid this by passing the size step3 needs directly from step2, for the common path (not implemented for change/delete conflicts for instance). I didn't fix the same race for the exec bit for now, because it's less likely to be problematic and I had trouble due to the fact that the dirstate stores the permissions differently from the manifest (st_mode vs '' 'l' 'x'), in combination with tests that pretend that symlinks are not supported. However, I moved the lstat from step3 to step2, which should tighten the race window markedly, both for the exec bit and for the mtime. Differential Revision: https://phab.mercurial-scm.org/D6475
-
- Jun 12, 2019
-
-
Valentin Gatien-Baron authored
Currently worker supports running functions that return a progress iterator. Generalize it to handle function that return a progress iterator then a return value. It's unused in this commit, but will be used in the next one. Differential Revision: https://phab.mercurial-scm.org/D6515
-
- May 19, 2019
-
-
sliquister authored
which can result in bad status output. Concretely, this seems to be easily triggered by having a build system watching the filesystem for changes, and rebuilding files that are both tracked and generated while an update is happening. Differential Revision: https://phab.mercurial-scm.org/D6474
-
- May 23, 2019
-
-
Georges Racinet authored
The --rust global option turns on usage (and by default compilation) of the rust-cpython based mercurial.rustext. Similarly to what's previously done for zstd, there is a --no-rust option for the build_ext subcommand in order not to build mercurial.rustext, allowing for an OS distribution to prebuild it. The HGWITHRUSTEXT environment variable is still honored, and has the same effect as before, but now it works mostly by making the --rust global option defaulting to True, with some special cases for the direct-ffi case (see more about that below) Coincidentally, the --rust flag can also be passed from the make commands, like actually all global options, in the PURE variable make local PURE=--rust This feels inappropriate, though, and we should follow up with a proper make variable for that case. Although the direct-ffi bindings aren't directly useful any more, we keep them at this stage because - they provide a short prototyping path for experiments in which a C extension module has to call into a Rust extension. The proper way of doing that would be to use capsules, and it's best to wait for our pull request onto rust-cpython for that: https://github.com/dgrunwald/rust-cpython/pull/169 - Build support for capsules defined in Rust will probably need to reuse some of what's currently in use for direct-ffi.
-
- May 30, 2019
-
-
Georges Racinet authored
This commit converts all current Python callers of mercurial.rustext to the new policy.importrust system. After this point, going through policy.importrust or policy.importmod (in some more distant future) is mandatory for callers of Rust code outside of Python tests. We felt it to be appropriate to keep Rust-specific tests run inconditionally if the Rust extensions are present.
-
- May 29, 2019
-
-
Georges Racinet authored
We introduce two rust+c module policies and a new `policy.importrust()` that makes use of them. This simple approach provides runtime switching of implementations, which is crucial for the performance measurements such as those Octobus does with ASV. It can also be useful for bug analysis. It also has the advantage of making conditionals in Rust callers more uniform, in particular abstracting over specifics like `demandimport` At this point, the build stays unchanged, with the rust-cpython based `rustext` module being built if HGWITHRUSTEXT=cpython. More transparency for the callers, i.e., just using `policy.importmod` would be a much longer term and riskier effort for the following reasons: 1. It would require to define common module boundaries for the three or four cases (pure, c, rust+ext, cffi) and that is premature with the Rust extension currently under heavy development in areas that are outside the scope of the C extensions. 2. It would imply internal API changes that are not currently wished, as the case of ancestors demonstrates. 3. The lack of data or property-like attributes (tp_member and tp_getset) in current `rust-cpython` makes it impossible to achieve direct transparent replacement of pure Python classes by Rust extension code, meaning that the caller sometimes has to be able to make adjustments or provide additional wrapping.
-
- Jun 13, 2019
-
-
Navaneeth Suresh authored
This patch adds an entry for `internals.mergestate` as suggested by @marmoute. Most of the help text is taken from `merge.mergestate`. Differential Revision: https://phab.mercurial-scm.org/D6448 Differential Revision: https://phab.mercurial-scm.org/D6528
-
- Jun 12, 2019
-
-
Ian Moody authored
Now that Mercurial's Phabricator instance has been updated to a version that supports the parents.set transaction on revision.edit we can use that to set dependency relationships in patch stacks instead of abusing the summary. This has the advantage that we can use it on every `phabsend` so commit reordering is picked up without spamming changes like abusing the summary would, and using parents.set will clear previous parents unlike parents.add. Differential Revision: https://phab.mercurial-scm.org/D6514
-
- May 31, 2019
-
-
amalloy authored
Specifically, the second 'with' in 'with which to merge with'. Differential Revision: https://phab.mercurial-scm.org/D6483
-
- Jun 10, 2019
-
-
Kyle Lippincott authored
One of our users was confused because they read this, and then attempted to run `hg rebase` with a dirty working directory, and it still complained. The reason was that they were attempting to rebase the commit they currently had checked out, which (at least with evolve workflows enabled) involves updating the working directory to be based on the newly rebased commit. Differential Revision: https://phab.mercurial-scm.org/D6507
-
Valentin Gatien-Baron authored
Currently, it is implemented on top of commonancestorsheads. Implement it on top of reachableroots instead, as reachableroots could stop walking the graph much sooner than commonancestorsheads. Measuring repo.changelog.isancestorrev on two revisions in a private repository: before: ! wall 0.005175 comb 0.010000 user 0.010000 sys 0.000000 (best of 550) after : ! wall 0.000072 comb 0.000000 user 0.000000 sys 0.000000 (best of 36199) When hg does this kind of operations 1500 times in pull -> bookmarks.comparebookmarks -> bookmarks.validdest, that's 11s that drop from the --profile output. Differential Revision: https://phab.mercurial-scm.org/D6506
-
Valentin Gatien-Baron authored
The previous revset couldn't be correct as it is symmetric in <roots> and <heads>, but reachableroots has no such symmetry. It makes a difference with for instance reachableroots(2, 3) where 2 and 3 are both children of 1. Differential Revision: https://phab.mercurial-scm.org/D6505
-
- Jun 11, 2019
-
-
Ian Moody authored
The way to signal to Conduit that a reviewer is considered blocking is just to wrap their PHID in "blocking()" when including it in the list of PHIDs passed to `reviewers.add`. arc doesn't have a --blocker, instead one is supposed to append a '!' to the end of reviewer names (I think reviewers are usually added in an editor rather than the command line, where '!'s can be more hazardous). moz-phab (Mozilla's arcanist wrapper) does have a --blocker argument, and being explicit like this is also more discoverable. Even `arc diff`'s help doesn't seem to mention the reviewer! syntax. Differential Revision: https://phab.mercurial-scm.org/D6512
-
Ian Moody authored
Currently when making VCR recordings one needs to manually sanitise sensitive credentials before committing and submitting them as part of tests. It is easy to imagine this being accidentally missed one time by a fallible human and said credentials being leaked. It is also possible that it wouldn't be noticed to alert the user to the leak since the recording files are so large and practically unreviewable. Thus do so automatically, so the only place that needs checking is in the test-phabricator.t file. Differential Revision: https://phab.mercurial-scm.org/D6513
-
Pulkit Goyal authored
Doing bytes[0] will return the ascii value of that position which breaks comparison with a bytechar. This makes test-absorb.t work again on py3. Differential Revision: https://phab.mercurial-scm.org/D6508
-
- Jun 09, 2019
-
-
Yuya Nishihara authored
I have a code which basically runs "0:wdir() & <user-revset>", and it crashed if merge() were passed in.
-
Yuya Nishihara authored
-
- Jun 01, 2019
-
-
Martin von Zweigbergk authored
Differential Revision: https://phab.mercurial-scm.org/D6470
-
- Jun 06, 2019
-
-
Raphaël Gomès authored
The current shortcut for rootglobs that can be simplified to exact matches does not work, it instead treats the pattern as a regex, which is not the same thing. This changes fixes the behavior and introduces a test for this behavior. Differential Revision: https://phab.mercurial-scm.org/D6489
-
Raphaël Gomès authored
In my initial patch, I introduced an unnecessary hard constraint on UTF-8 filenames and patterns which I forgot to remove. Although the performance penalty for using String might be negligible, we don't want to break compatibility with non-UTF-8 encodings for no reason. Moreover, this change allows for a cleaner Rust core API. This patch introduces a new utils module that is used with this fix. Finally, PatternError was not put inside the Python module generated by Rust, which would have raised a NameError. Differential Revision: https://phab.mercurial-scm.org/D6485
-
- May 31, 2019
-
-
Jörg Sonnenberger authored
Differential Revision: https://phab.mercurial-scm.org/D6467
-
- Jun 08, 2019
-
-
Pulkit Goyal authored
I am trying to write some automations around phabricator and having debugcallconduit work outside a hg repo will be nice! Marking command as optionalrepo instead of norepo because we might to load repo/.hg/hgrc. Differential Revision: https://phab.mercurial-scm.org/D6499
-
Pulkit Goyal authored
This will help us make `hg debugcallconduit` work outside a hg repo as next patch will mark that command as no repo. Differential Revision: https://phab.mercurial-scm.org/D6498
-
Pulkit Goyal authored
The goal of this series is to make `hg debugcallconduit` work outside of a hg repo. This patch, removes requirement of repo object from readurltoken as we only need ui there. It also updates the callers to pass in ui instead of repo. Differential Revision: https://phab.mercurial-scm.org/D6497
-
Pulkit Goyal authored
I installed emacs on the server running buildbot and the test started passing on Python 3. Lets add it to the list of passing test. Differential Revision: https://phab.mercurial-scm.org/D6500
-
- Jun 07, 2019
-
-
Ian Moody authored
Especially useful when sending updates to existing Revisions so one can specify the sort of changes e.g. "Address review comments" or "Rebase to tip" If the diff content hasn't changed then it only needs a metadata update which doesn't show in the Phabricator updates UI, so don't add a comment that will. Differential Revision: https://phab.mercurial-scm.org/D6496
-
- Jun 05, 2019
-
-
Pulkit Goyal authored
For some reasons, the output of print was not going through. Replaced that ui.status(). Differential Revision: https://phab.mercurial-scm.org/D6481
-
Pulkit Goyal authored
# skip-blame as just b'' prefixes Differential Revision: https://phab.mercurial-scm.org/D6480
-
Pulkit Goyal authored
#skip-blame because just b'' prefix Differential Revision: https://phab.mercurial-scm.org/D6482
-
- Jun 06, 2019
-
-
Martin von Zweigbergk authored
This makes it more consistent with how we encode copies (newline-separated lists of null-separated pairs). This perhaps makes {extras} a little less readable (?) despite avoiding the escaping. I don't know how I feel about this patch. I'm okay with it being queued or dropped. Differential Revision: https://phab.mercurial-scm.org/D6486
-