branchcache: use update_disk to refresh 'served' and 'served.hidden' The `update_disk` method is dedicated to this kind of usecase. Now that the writting patterns are more consistent, we can use it to warm these two important cache. I am dropping the first comment about "refreshing all the others" because it is false. If a branchmap already exist for "served", none of the subset will be updated.

branchcache: explictly update disk state only if no transaction exist If a transaction exist the `write_dirty` call will eventually be done and the state will be synched on disk. It is better to no interfer with that.

branchcache: do not use `__getitem__` in updatecache The `update_disk` method uses `updatecache` and the point of `update_disk` is to be able to do alternative processing to the one we do in `__getitem__`. So we calling `__getitem__` in `updatecache` defeat this purpose. Instead we do the equivalent explicitly to preserve the spirit of `update_disk` (that we will actually put to use soon, I promise)

branchcache: explicitly track inheritence "state" We move from a binary "dirty" flag to a three value "state": "clean", "inherited", "dirty". The "inherited" means that the branch cache is not only "clean", but it is a duplicate of its parent filter. If a branch cache is "inherited", we can non only skip writing its value on disk, but it is a good idea to delete any stale value on disk, as those will just waste time (and possibly induce bug) in the future. We only do this in the update related to transaction or explicit cache update (e.g `hg debugupdatecache`). Deleting the file when we simply detected a stall cache during a read only operation seems more dangerous. We rename `copy` to `inherit_for` to clarify we associate a stronger semantic to the operation.

branchcache: stop writing more branchcache file on disk than needed Before this change, we were unconditionally writing a branchmap file for the filter level passed to `update_disk`. This is actually counter productive if no update were needed for this filter level. In many case, the branch cache for a filter level is identical to its parent "subset" and it is better to simply keep the subset update and reuse it every time instead of having to do identical work for similar subset. So we change the `update_disk` method to only write a file when that filter level differ from its parent. This removes many cases where identical files were written, requiring multiple boring update in the test suite. The only notable changes is the change to `test-strip-branch-cache.t`, this case was checking a scenario that no longer reproduce the bug as writing less branchmap file result in less stalled cache on disk. Strictly speaking, we could create a more convoluted scenario that create a similar issue. However the next changeset would also cover that scenario so we directly updated that test case to a "no longer buggy" state.

branchcache: do not copy the `_dirty` flag If the inherited branch cache is dirty, it will be written on disk, and the super-set did not need to modify it, the on disk value for the subset will be re-useable as is. So the super set does not needs to write the very same content itself.

branchcache: explicitly assert that copy is always about inheritance This would catch cases where copy is used for something else if any existed.

branchcache: stop using `copy(…)` in `replace(…)` The `copy` method is mostly used for a filter level to inherit the branchmap from a subset. So we stop using (abusing) it in "replace" to ensure `copy` is used only for inheritance purposes. Since `replace` is a method of the BranchMapCache, it seems fine to do lower level operation there.

branchcache: change the _delayed flag to an explicit `_dirty` flag This is more consistent with the logic we use for other object and it open the way to a clearer management of the cache state. Now, cache are created clean, cache update mark them dirty, writing them on disk mark them clean again.

branchcache: write branchmap in subset inheritance order This way, we can guarantee a valid subset has been written before touching the branchmap of another filter. This is especially useful as we are bout to start deleting outdated branchmap file.

branchcache: do not accept "empty update" This currently does not happens and it will be simpler that is remains that way. If all update do something, we will be able to simply declare, in a later changesets, that all update to result in a dirty branchcache.

branchcache: avoid created a `None` filter repoview when writing The repoview class is not intended to be used for unfiltered repository.

stream-clone-tests: stop filtering non existent warning This filtering was introduced in 74c004a515bc, however there is already no warning in that changeset. So I guess the warnings existed when we the patch was created but the problem was solved in another changeset that 74c004a515bc, rebased on.

stream-clone-test: simplify case testing obsolescence There is only two important things in this test: - the number of file we send, to show we picked the obsstore. - the resulting state, to show we did alter things in the process. The rest are of the number are very fragile and consume a lot of time for little value when adjusting formats, caches, and protocol.

stream-clone-test: simplify the case testing phases There is only two important things in this test: - the number of file we send, to show we picked the phase roots. - the resulting phases, to show we did not modified them. The rest are of the number are very fragile and consume a lot of time for little value when adjusting formats, caches, and protocol.

stream-clone-test: simplify bookmark clone The important things to test here is the number of file included (to catch that the bookmark file was sent). So we keep that part non glob'ed but glob the rest. The glob'ed numbers are very fragile and consume a lot of time for little value when adjusting formats, caches, and protocol.

stream-clone-test: add a verify call to the "clone while changing" case It seems useful to very that the clone did not result in a corrupted copy.

stream-clone-test: add title to various test cases These case are fine as is, but as we are adding title to all the other as we simplify them, lets add title for all cases.

stream-clone-test: simplify testing of secret cloning restriction Here, we just want to check if the streaming clone is allowed and used or not. We do not care about the details of the clone itself.

stream-clone-test: simplify the background file closing test Here we just care about the fact the background file closing logic actually ran. We don't need to check the details of the cloning. The details of the output is very fragile and consume a lot of time for little value when adjusting formats, caches, and protocol. So we filter it out.

stream-clone-test: simplify the --uncompressed alias check To check that --uncompressed is an alias we just need to check it trigger a stream clone, we don't need to check anything else.

stream-clone-test: drop an automatic pattern replacement That pattern is nice, but it prevent us to glob the number of bytes when we don't care about them. We don't care about them more often that what we currently checks so dropping this pattern will help use to simplify various tests.

stream-clone-test: simplify the test for getbundle with stream=1 The core of this tests is about checking we receive a stream bundle with such request. We don't need to look at too much of the details of the stream itself. Since the content of the stream if shifting overtime, Such check is very fragile and consume a lot of time for little value when adjusting formats, caches, and protocol. So we reduce the size of what we check to focus on "is this a stream clone" question.

stream-clone-test: factor some piece of basic clone test out Multiple parts of this case (listing cache, checking error) are common to all cases and don't need to be in the conditionnal block. This simplify the test update.

stream-clone-test: simplify the case where server disabled it We have an option to disable it, we don't need to test it with all protocol variants. In addition there is little value in looking at the bytes to bytes details of the reply. Such check is very fragile and consume a lot of time for little value when adjusting formats, caches, and protocol.

repoview: prevent `None` to be passed as the filtername We let such instantiation slip in a previous commit, so we add an explicit check to prevent it to happen in the future.

branchcache: avoid created a `None` filter repoview when writing The repoview class is not intended to be used for unfiltered repository.

branchcache: move head writing in a `_write_headers` method Same rational: this will help having format variants.

branchcache: move head writing in a `_write_heads` method Same rational: this will help having format variants.

branchcache: move the header loading in a `_load_header` class method This will help changing header parsing in format variants.

branchcache: simplify a long line Gratuitous change to help code readability.

branchcache: rename `load` to `_load_heads` We are about to have more similar function, we rename the existing one to a more meaningful name and mark it private in the process.

branchcache: move the filename to a class attribute This prepare the introduction of more variant of cache.

test-clonebundles: simplify matching to be less flavor depends We keep the files and bytes output for the first call, but then we mostly check that we are being served a stream-clone bundle, not the actual content and size of the bundle. That aspect being tested by the stream clone test themselves.

repoview: fix changelog.__contains__ method This have been around for ten years, so we can safely that this method have few callers. However I am about to add one.

branchcache: unconditionally write delayed branchmap

branchcache: drop the unused `_verifyclosed` This code appears dead since its introduction about 5 years ago in this three consecutive commits: - 6578654916ae → introduce the method with two calls - 7c9d4cf23adf → remove first call - be5eeaf5c24a → remove second call o changeset: be5eeaf5c24a | user: Pulkit Goyal <pulkit@yandex-team.ru> | date: Fri Apr 05 15:57:09 2019 +0300 | summary: branchcache: don't verify closed nodes in _branchtip() | o changeset: 7c9d4cf23adf | user: Pulkit Goyal <pulkit@yandex-team.ru> | date: Fri Apr 05 15:56:33 2019 +0300 | summary: branchcache: don't verify closed nodes in iteropen() | o changeset: 6578654916ae | user: Pulkit Goyal <pulkit@yandex-team.ru> ~ date: Mon Apr 01 13:56:47 2019 +0300 summary: branchcache: lazily validate nodes from the branchmap

branchcache: dispatch the code into the dedicated subclass The code useful only to the local brancache have now been moved into the dedicated subclass. This will help improving the branchcache code without subtle breaking the remote variants.

branchcache: introduce a base class for branchmap This will help define a clear boundary between the two.

branchcache: fix the copy code We copy some internal attribute along too. This should prevent inconsistency in the resulting branchmap.

branchcache: pass a "verify_node" attribut to __init__ instead of hasnode The hasnode callback cannot be inherited and is dropped on copy, which seems like a bad idea. Instead we pass the actual semantic as a parameter and let the internal logic deal with it.

branchcache: stop storing a repository instance on the cache altogether We did not really needed it and we do not needs it anymore at all. So lets make things simpler for consistency and garbage collecting and stop storing it altogether.

branchcache: pass the target repository when copying Branchmap are usually copied to be used on a different repoview using a different filter level. Passing the repository around means the repository in `branchcache._repo` will drift from the actual branchmap filter. This is currently "fine" because the repo is only used to retrieve the `nullid` value. However, this is a fairly big trap for any extension or future code using the `_repo` attribute. The replace logic is now using a copy to ensure the right repository view is used to initialized the cached value. We add a couple of assert for make sure this inconsistency does not sneak back.

branchcache: have an explicit method to update the on disk cache Explicit is better and will give use more flexibility for future evolution of the storage.

rust-index: drop offset_override The inline `offsets` value diverge from the one on disk for added value, so the offset_override tricks is not going to work well once we start having the full revlog logic in Rust. We remove it beforehand and align the Rust logic to the Python one (adjusting the segment offset at read time for inline revlog).

rust-index: stop calling `with_offset` in the tests We are not adding any data, so why are we setting any offset?

branching: merge stable into default

test: better glob some timing related line to avoid flakiness If we go over 10 seconds, the number of white space changes.

Added signature for changeset a1a011d4b148

Added tag 6.7.4 for changeset a1a011d4b148

relnotes: add 6.7.4 and warn about 6.7.{1,2,3}

inline-changelog: fix pending transaction visibility when splitting We move the name back to the expected name of `changelog.i.a`.

inline-changelog: fix a critical bug in write_pending that delete data Since a93e52f0b6ff we no longer use inline-revlog for the changelog. The goal there was to solve the lack of testing for the two variants (inline vs split) and reduce the complexity of the interaction with "diverted-write" on the changelog level. However many existing repository still have inline-changelog and we automatically move them to normal revlog as soon as we have the chances. Unfortunately This conversion is buggy and can result in the destruction of the changelog.i if hook triggers the "write pending" mechanism. The bugs comes from the "revlog splitting" logic and the "write_pending" logic stepping over each other. Ironically the change in a93e52f0b6ff aims at no longer having this kind of problem. This changesets fix this issue and add associated tests. Fixing this reveal that the transaction hooks end up not seeing the pending transaction content, because the name is not right ("changelog.i.s.a" instead of "changelog.i.s") we fix this in the next changeset.

bookmark: fix remote bookmark deletion when the push is raced Before this patch, running `hg push -B book` to push the `book` bookmark sideway at the same time as a commit making it moving forward might result in the removal of the bookmark remotely. After this changeset, the push can still be raced, but to remove deletion happens. This is progress.

hooks: add a prewlock and a prelock hooks This is useful for testing.

exchange: fix locking to actually be scoped The previous code was taking locks before entering with statements, so exception before the with statement would not release the lock (except for garbage collection). We need to move to a try except here because the logic is more complicated.

exchange: fix locking to actually be scoped The previous code was taking locks before entering with statements, so exception before the with statement would not release the lock (except for garbage collection).

rust-status: sort the failed matches when printing them This was making the tests flaky after the recent patch¹ that opened up more of the code to the Rust-augmented status. [1] 865efc020c3355dca1cbaa35db80600009c01dd5

chistedit: change action for the correct item We have an experimental config histedit.later-commits-first from c820866c52f9, and when it's true, the order of commits in histedit UI is reversed, both in text mode and in curses mode. But before this patch key presses in curses mode would change histedit actions in the same old order, i.e. trying to edit the latest commit (which would be first now) would put "edit" action on the last commit in the list. This wasn't a cosmetic issue, histedit would actually proceed to edit the first commit in the list. Let's map rules to display items (hopefully now correctly).

Added signature for changeset 028dc3f92dbd

Added tag 6.7.3 for changeset 028dc3f92dbd

relnotes: add 6.7.3

rust: blanket implementation of Graph for Graph references The need comes from the fact that `AncestorsIterator` and many Graph-related algorithms take ownership of the `Graph` they work with. This, in turn is due to them needing to accept the `Index` instances that are provided by the Python layers (that neither rhg nor `RHGitaly` use, of course): the fact that nowadays the Python layer holds an object that is itself implemented in Rust does not change the core problem that they cannot be tracked by the borrow checker. Even though it looks like cloning `Changelog` would be cheap, it seems hard to guarantee that on the long run. The object is already too rich for us to be comfortable with it, when using references is the most natural and guaranteed way of proceeding. The added test seems a bit superfleous, but it will act as a reminder that this feature is really useful until something in the Mercurial code base actually uses it.

rust-cpython: don't swallow the dirstate error message In case we do get a dirstate error, we want to get the full error message and not just an opaque `Dirstate error`.

dirstate-v2: check that root nodes are at the root before writing More explanations in the previous changeset.

dirstate-v2: add check of parent/child nodes being related when writing This stems from a corruption seen in a private repository. We're not sure of the source of the corruption, and it's very possible that we're seeing compounded effects of multiple writes on a corrupted dirstate. Adding this check is not expensive in itself and large writes of the dirstate are not common. This change does not catch this problem at the root node, the next one will.

admin-verify: expect a number of errors to be returned It's the responsibility of the check to handle errors, we only care about the total count to sum up the check's work. We use `admin::verify -c dirstate` to test this path at least somewhat.

admin-verify: fix error message handling `dirstate.verify` used to return tuples but does not anymore, it returns the pre-formatted error message, which is a nicer interface anyway.

admin-verify: pass p1 down to the dirstate function This was forgotten and can break with certain kinds of corruption.

Backed out changeset 3e0f86f09f26

Backed out changeset fc317bd5b637

re2: make errors quiet By default, the re2 library will output error on its own instead of keeping the error in an exception. This make re2 printing spurious error before fallback to the stdlib remodule that may accept the pattern or also fails to parse it and raise a proper error that will be handled by Mercurial. So we also pass an Option object that changes this default.

fold-or-prune-me: update proposal This does the same things but with a narrower wrapping.

subrepo: propagate non-default path on outgoing There was already a fix made in 5dbff89cf107 for pull and push commands. I did the same for the outgoing command. The problem I identified is that when the parent repository has multiple paths, the outgoing command was not respecting the parent path used and was always using the default path for subrepositories.

wireprotoserver: ensure that output stream gets flushed on exception Previously flush was happening due to Python finalizer being run on `BufferedWriter`. With upgrade to Python 3.11 this started randomly failing. My guess is that the finalizer on the raw `FileIO` object may be running before the finalizer of `BufferedWriter` has a chance to run. At any rate, since we're not relying on finalizers in the happy case we should also not rely on them in case of exception.

match: strengthen visit_children_set invariant, Recursive means "all files" My previous interpretation of "Recursive" was too relaxed: I thought it instructed the caller to do something like this: > you can stop calling `visit_children_set` because you'll need to descend into > every directory recursively, but you should still check every file if it > matches or not Whereas the real instruction seems to be: > I guarantee that everything in this subtree matches, you can stop > querying the matcher for all files and dirs altogether. The evidence to support this: - the test actually passes with the stronger invariant, revealing no exceptions from this rule - the implementation of `visit_children_set` for `DifferenceMatcher` clearly relies on this requirement, so it must hold for that not to lead to bugs.

match: fix the rust-side bug in visit_children_set for rootfilesin matchers The fix is checked by `test_pattern_matcher_visit_children_set` test, which is what caught the bug in the first place, but also by an end-to-end test that I made for this purpose. Accept the new results of Cargo tests Many of these were already annotated with "FIXME", which is a good sign.

match: fix the "visitdir" method on "rootfilesin" matchers This fixes just the Python side, the fix for the rust side will follow shortly.

match: rename RootFiles to RootFilesIn for more consistency

match: small tweak to PatternMatcher.visit_children_set This makes it a bit more efficient (avoid a computation in case of early return), and in my opinion clearer.

matchers: fix the bug in rust PatternMatcher that made it cut off early This brings the rust output in line with the Python output.

tests: add an end-to-end test to show a bug in `visit_children_set` Concretely, `rootfilesin` is completely broken with respect to `visit_children_set` optimization.

tests: add tests and document expectations from visit_children_set in rust The tests this patch are adding have the form of formal spec in invariants::visit_children_set::holds, and then a series of checks that all examples must satisfy this formal spec. I tried to make the spec consistent with how this function is used and how it was originally conceived. This is in conflict with how it's documented in Rust. Some of the implementations also fail to implement this spec, which leads to bugs, in particular when complicated patterns are used with `hg status`.

tests: add a test that demonstrates a bug in rhg status pattern handling The bug is in [visit_children_set], will be elaborated on in follow-up changes.

bundle-spec: properly parse boolean configuration as boolean Before this changesets "v2;revbranchcache=no" would actually request the addition for a revbranchcache part as the non-empty string `"0"` is `True`

bundle-spec: properly identify changegroup-less bundle It is possible to produce a bundle without changegroup. For example if we want to only send phases or obsolescence information. However that lead to crash for command that identifies bundle content. So we fix that. The test will come in the next changesets, when we fix another bug preventing to generate such bundle by hand.

setup: display return code information about failed `hg` call This help to understand what is going wrong when things goes wrong.

Added signature for changeset 803e61387e86

Added tag 6.7.2 for changeset 803e61387e86

relnotes: add 6.7.2

bundle2: make the "hgtagsfnodes" part advisory This bundle2 part is about helping the client to warms its cache. There is no reason for it to be mandatory. So we mark it advisory.

phases: update the phase set as we go during retract boundary Apparently iterating over the `changed_revs` dictionary is very expensive. On mozilla-try-2019-02-18, a perf::unbundle call with a 10 000 changesets bundle gives give use the following timing. e57d4b868a3e: 4.6 seconds ac1c75188440: 102.5 seconds prev-changeset: 30.0 seconds this-changeset: 4.6 seconds So, the performance regression is gone. Once again: thanks to marvelous Python!

phases: avoid a potentially costly dictionary interation in some case If we retract for the draft phase, there is not non-public item to be retracted and we can skip this part. This part is was apparently super costly thanks to Python. On mozilla-try-2019-02-18, a perf::unbundle call with a 10 000 changesets bundle gives give use the following timing. e57d4b868a3e: 4.6 seconds ac1c75188440: 102.5 seconds this-changeset: 30.0 seconds So we recovered about ⅔ of the regression, the next changeset will give us the rest back.

Added signature for changeset 2e6fde2ed01e

Added tag 6.7.1 for changeset 2e6fde2ed01e

relnotes: add 6.7.1

subrepo: fix normalizing paths with scheme After revision 0afe96e374a7, subrepo paths were normalized using posixpath.normpath and that resulted in ssh paths being wrongly converted from ssh://host/path to ssh:/host/path This fix applies the same logic used in urlutil.url to split the path scheme from the rest and only use posixpath.normpath to the string after scheme://

sshpeer: fix path when handling invalid url exception In 73ed1d13c0bf the code was refactored but the error handling seems to have been missed (or maybe the object shoud have implemented __bytes__)

delta-search: fix crash caused by unbound variable This code path was apparently not tested. This fixes a crash when cloning the Tryton repo.

Added signature for changeset c9ceb4f60256

Added tag 6.7 for changeset c9ceb4f60256

phases: avoid N² behavior in `advanceboundary` We allowed duplicated entries in the deque, which each entry could potentially insert all its ancestors. So advancing boundary for the full repository would mean each revision would walk all its ancestors, resulting in O(N²) iteration. For repository of any decent size, N² is quickly insane. We introduce a simple set to avoid this and get back to reasonable performance.

relnotes: add 6.7

admin-commands: move the chainsaw extension to the admin commands module Activating an extension is always a little bit of a chore and the long name, options and "chainsaw" bits are deterrent enough. This also allows us to help the discoverability for people looking for repo "administration" tools, with the widest semantic of "administration".

obsutil: sort metadata before comparing in geteffectflag() This is probably less important now that we dropped Python 2. We do still support Python 3.6 though, and the dictionaries aren't ordered there either (that was a big change that came with 3.7). Still, maybe it's a good idea to sort metadata explicitly.

tests: disable revlog compression in test-generaldelta.t (issue6867) The revlog compression makes a lot of numbers unstable. Since checking revlog compression is not the goal of this test, we disable the compression to get stable numbers. This should avoid wasting more time on this kind of changes in the future.

test-general-delta: actually test optimize-delta-parent-choice=no Since the configuration was not explicit, the case stopped testing what it intended to test when the default value changed.

test-chg: stabilize the log checking The "worker process exited" line have been making the CI flaky for a long time. Lets sort this out.

tests: fix test-patchbomb-tls.t instability The flakiness on chg is caused by a client that exit faster than the server output log. So actively wait for the server to issue the expected output (with a small timeout)

test-lock: use synchronisation file instead of sleep This will prevent the test to be flaky on load.

rust-matchers: raw regular expression builder Extracting this `re_builder()` from `re_matcher()` makes it reusable in more general cases than matching `HgPath` instances and would help reducing code duplication in RHGitaly.

rust-filepatterns: export glob_to_re function Making this function public should not risk freezing the internal API, and it can be useful for all downstream code that needs to perform glob matching against byte strings, such as RHGitaly where it will be useful to match on branches and tags.

rust-index: don't use mutable borrow to computed filtered heads This does not need to mutate the index. This is the prime suspect for some RuntimeError raised during some pushes.

rust-index: don't use mutable borrow for head-diff computation It does not needs to mutate the index. This is one of the two suspects of RuntimeError being thrown during push.

crecord: drop calls to `curses.endwin()` We got a bug report where `curses.endwin()` failed with `_curses.error: endwin() returned ERR`. Looking at e306d552dfb12, it seems like we should be able to just remove these calls.

config: move the option to mmap rev branch cache in the storage section See previous commit for rational.

config: document the storage and format sections This should help people to put configuration in the right section.

Added signature for changeset d1d48d18db37

Added tag 6.7rc0 for changeset d1d48d18db37

relnotes: add 6.7rc0

relnotes: remove outdated message from `next`

branching: merge default into stable for 6.7rc0

branching: merge stable into default

perf: add a --as-push option to perf::unbundle This turned out to make a quite significant difference.

chainsaw-update: exit early if one of the intermediate command fails That will prevent the user to be presented with a start that pretend to be consistent with the request, but is not.

chainsaw-update: lock the repository for the duration of the operation This should prevent and catch some misusage where something else try to touch the repository.

chainsaw-update: taking care of initial cloning Perhaps we should go just a bit lower level than this `instance()`, since the main added value in our use-case is full path resolution, that we need to do anyway for the rmtree cleanup.

chainsaw-update: use a graph with branching in graph This will be relevant for the next improvement of `chainsaw-update`.

chainsaw-update: log actual locks breaking Previously, the command would simply state that it was about to break locks, not if there was actually some to break. This version is race-free. It would be also possible to display the content of the lock before hand (not race-free but informative in almost all cases).

vfs: have tryunlink tell what it did It is useful in certain circumstances to know whether vfs.tryunlink() actually removed something or not, be it for logging purposes.

chainsaw: new extension for dangerous operations The first provided command is `chainsaw-update`, whose one and single job is to make sure that it will pull, update and purge the target repository, no matter what may be in the way (locks, notably), see docstring for rationale.

rust: disable the RustIndex without persistent nodemap See rational inline.

rust: stop claiming the C index is compatible with the rust code This is no longer the case since the introduction of the pure Rust Index, and was probably not the case since the MixedIndex itself. So we fix the dedicated attribute value.

rust-index: remove one collect when converting back Turns out this is slightly faster. Sending the results back to Python is still the most costly (like 75% of the time) of the whole method, but it's about as fast as it can be now. hg perf::phases on mozilla-try-2023-03-22 before: 0.267114 after: 0.247101

rust-index: improve phase computation speed While less memory efficient, using an array is *much* faster than using a HashMap, especially with the default hasher. It even makes the code simpler, so I'm not really sure what I was thinking in the first place, maybe it's more obvious now. This fix a significant performance regression when using the rust version of the code. (however, the C code still outperform rust on this operation) hg perf::phases on mozilla-try-2023-03-22 - 6.6.3: 0.451239 seconds - before: 0.982495 seconds - after: 0.265347 seconds - C code: 0.183241 second

phases: directly update the phase sets in advanceboundary This is similar to what we do in retractboundary. There is no need to invalidate the cache if we have everything at hand to update it.

phases: large rework of advance boundary In a similar spirit as the rework of retractboundary, the new algorithm is doing an amount of work in the order of magnitude of the amount of changeset that changes phases. (except to find new roots in impacted higher phases if any may exists). This result in a very significant speedup for repository with many old draft like mozilla try. runtime of perf:unbundle for a bundle constaining a single changeset (C code): before 6.7 phase work: 14.497 seconds before this change: 6.311 seconds (-55%) with this change: 2.240 seconds (-85%) Combined with the other patches that fixes the phases computation in the Rust index, the rust code with a persistent nodemap get back to quite interresting performances with 2.026 seconds for the same operation, about 10% faster than the C code.

phases: apply similar early filtering to advanceboundary advanceboundary is called the push's unbundle (but not the other unbundle) so advanceboundary did not show up the profile I looked at so far. We start with simple pre-filtering to avoid doing any work if we don't needs too.

phases: filter revision that are already in the right phase No need to compute new roots if everything is already in order.

phases: invalidate the phases set less often on retract boundary We already have the information to update the phase set, so we do so directly instead of invalidating the cache. This show a sizeable speedup in our `perf::unbundle` benchmark on the many-draft mozilla-try repository. ### data-env-vars.name = mozilla-try-2023-03-22-zstd-sparse-revlog # benchmark.name = hg.perf.perf-unbundle # bin-env-vars.hg.flavor = no-rust # bin-env-vars.hg.py-re2-module = default # benchmark.variants.issue6528 = disabled # benchmark.variants.revs = last-10 before: 2.055259 seconds after: 1.887064 seconds (-8.18%) # benchmark.variants.revs = last-100 before: 2.409239 seconds after: 2.222429 seconds (-7.75%) # benchmark.variants.revs = last-1000 before: 3.945648 seconds after: 3.762480 seconds (-4.64%)

phases: incrementally update the phase sets when reasonable When the amount of manual walking is small, we update the phases set manually instead of computing them from scratch. This should help small update. The next changesets will make this used more often by reducing the amount of full invalidation we do on roots upgrade. The criteria for using an incremental upgrade are arbitrary, however, it "should never hurt".

phasees: properly shallow caopy the phase sets dictionary We are about to increments the set more incrementally in some case, so we need to make a proper shallow copy of it.

phases: pass an unfiltered repository to _ensure_phase_sets It seems better for such a low level function to be able to assume it operate on a real repository.

phases: drop set building in `hasnonpublicphases` We don't actually use the set, so why do we ensure they are built? (we should also clean up the use of repository argument but that's a quest for later).

phases: gather the logic for phasesets update in a single method This logic is duplicated around for no good reason, we gather it in a single place. The conditional is the new function are a bit weird as we about going to extend it soon.

phases: change the way we warm the phasecache in repocache Same logic as for the previous chngeset. We are about to rename and change the method used here.

phases: use a more generic way to trigger a phases computation for perf Querying the tip most revision will require the cache to warm the same as calling the dedicated method. This avoid using a method that is mostly meant for internal use and will be renamed in a coming changesets.

phases: fix an overzealous invalidation of the phase sets If `len(cl) == self._loadedrevslen` the cache is up to date.

phases: type annotation for `_phasesets` Does not hurt.

phases: leverage the collected information to record phase update Since the lower level function already gather this information, we can directly use it. This comes with a small change to the test that are actually fixing them. The previous version over-reported some phase change that did not exists. In both case, we are force revision `1` to be secret and `0` remains draft`, the previous code wrongly reported `0` as moving to secret while it properly remained draft in the repository.

phases: large rewrite on retract boundary The new code is still pure Python, so we still have room to going significantly faster. However its complexity of the complex part is `O(|[min_new_draft, tip]|)` instead of `O(|[min_draft, tip]|` which should help tremendously one repository with old draft (like mercurial-devel or mozilla-try). This is especially useful as the most common "retract boundary" operation happens when we commit/rewrite new drafts or when we push new draft to a non-publishing server. In this case, the smallest new_revs is very close to the tip and there is very few work to do. A few smaller optimisation could be done for these cases and will be introduced in later changesets. We still have iterate over large sets of roots, but this is already a great improvement for a very small amount of work. We gather information on the affected changeset as we go as we can put it to use in the next changesets. This extra data collection might slowdown the `register_new` case a bit, however for register_new, it should not really matters. The set of new nodes is either small, so the impact is negligible, or the set of new nodes is large, and the amount of work to do to had them will dominate the overhead the collecting information in `changed_revs`. As this new code compute the changes on the fly, it unlock other interesting improvement to be done in later changeset.

phases: fast path public phase advance when everything is public Everything is already public, so we have nothing to do here.

phases: fast path retract of public phase There are no boundary to retract, so lets do nothing.

phases: keep internal state as rev-num instead of node-id Node-id are expensive to work with, dealing with revision is much simple and faster. The fact we still used node-id here shows how few effort have been put into making the phase logic fast. We tend to no longer use node-id internally for about ten years. This has a large impact of repository with many draft roots. For example this Mozilla-try copy have ½ Million draft roots and `perf::unbundle` see a significant improvement. ### data-env-vars.name = mozilla-try-2023-03-22-zstd-sparse-revlog # benchmark.name = hg.perf.perf-unbundle # bin-env-vars.hg.flavor = no-rust # bin-env-vars.hg.py-re2-module = default # benchmark.variants.issue6528 = disabled # benchmark.variants.revs = last-1 before:: 1.746791 seconds after:: 1.278379 seconds (-26.82%) # benchmark.variants.revs = last-10 before:: 3.145774 seconds after:: 2.103735 seconds (-33.13%) # benchmark.variants.revs = last-100 before:: 3.487635 seconds after:: 2.446749 seconds (-29.85%) # benchmark.variants.revs = last-1000 before:: 5.007568 seconds after:: 3.989923 seconds (-20.32%)

phases: do filtering at read time This remove the need for the `filterunknown` method at all.

phases: always write with a repo In the future change that move the internal representation of phase-roots from node-id to rev-num, we will use a repository to translate revision numbers back to node at write time. Since that future change is quite complicated already, we do this small API change beforehand.

phases: mark `phasecache.phaseroots` private We are about to change its content from nodeid to revnum. So anyone directly using the content might be in unexpected troubles. We start by making it private to explicitly break any such user (and discourage them to do so).

phases: check secret presence the right way during discovery There is an official function for this, lets use it. This will prevent the code to break in the future while we refactor the phase code.

phases: explicitly filter stripped revision at strip time Explicit is better than implicit. The current logic is bit subtle and fragile. It also get in the way of using something else than node-id as internal storage. We replace it with a more explicit filtering while striping.

debug: add a debug::unbundle command that simulate the unbundle from a push The code have different behavior when the unbundle comes from a push, so we introduce a command that can simulate such unbundle. For our copy of mozilla-try-2023-03-22, this make the unbundle jump from 2.5 seconds (with `hg unbundle`) to 15 seconds (with `hg debug::unbundle`). That 15 seconds timings is consistent with the issue seen in production.

annotate: limit output to range of lines

revlog: add a Rust implementation of `headrevsdiff` Python implementation of `headrevsdiff` can be very slow in the worst case compared with the `heads` computation it replaces, since the latter is done in Rust. Even the average case of this Python implementation is still noticeable in the profiles. This patch makes the computation much much faster by doing it in Rust.

revlog: add a C implementation of `headrevsdiff` Python implementation of `headrevsdiff` can be very slow in the worst case compared with the `heads` computation it replaces, since the latter is done in C. Even the average case of this Python implementation is still noticeable in the profiles. This patch makes the computation much much faster by doing it in C.

unbundle: faster computation of changed heads To compute the set of changed heads it's sufficient to look at the recent commits, instead of looking at all heads currently in existence.

branching: merge stable into default

hg-core: separate timestamp and extra methods

hg-core: implement timestamp line parsing

doc: document that labels must have a dot in them to have an effect I noticed that the `hg topics` template has a bare `topic` label with no dot, and that makes it useless, as such a label will never receive any effect by the colour extension. This dot has been required for a long time, at least since 2011, but we never formally documented it!

crecord: enable search hotkeys (issue6834) The keys I chose here should be similar to less/vim keybindings, which should fit the overall keybinding theme of crecord.

crecord: add handle(next|prev)search functions These are now just simple wrappers around `searchdirection`

crecord: add a searchdirection function If a regex has already been previously set, this function handles the UI elements of searching again forward or backward.

crecord: add a handlesearch function This function sets up some of the UI, such as getting the search string from the user and displaying results or their absence.

crecord: add a showsearch function This function takes a regex and searches either forward or backward, moving the current item to the found item, if any, and unfolding the relevant context.

crecord: add a default regex to curseschunkselector Whether there is a regex to search or not will affect if we can find the next or the previous search hit.

crecord: add `content` properties to all nodes In order to have a unified API of what can be searched, let's provide a `content` property to each node type. This way we can search filenames, context headers (e.g. containing function names, if deducible from patch context) or changed lines themselves.

crecord: update uiheader docstring There's no need to move anything to patch.py. The uiheader class only has methods relevant to crecord and overrides __getattr__ in order to use `patch.header` objects as a sort of mixin.

crecord: add skipfolded param to previtem This just simplifies the API a bit so it matches `nextitem` and I can handle both nextitem and previtem symmetrically.

dispatch: don't attempt to import debugger as bytestring The __import__ thingie needs a string, not a bytestring. Guess I'm the only one who uses this once in a while and noticed it was broken.

debugsetparents: fix Marmoutian docstring Just some light proofreading.

docs: fix broken `make` in `docs/` We had some wrapped lines without blank lines between, which made the runrst script think the list was not a list and it got confused about the indentation. I added blank lines, and also some other minor styling for consistency with the rest of the file.

branchmap: use mmap for faster revbranchcache loading A typical revbranchmap usage is: - load the entire revbranchmap file into memory - maybe do a few lookups - add a few bytes to it - write the addition to disk There's no reason to load the entire revbranchmap into memory. We can split it into a large immutable prefix and a mutable suffix, and then memorymap the prefix, thus saving all the useless loading. Benchmarking on some real-world pushes suggests that out of ~100s server-side push handling revbranchcache handling is responsible for: * ~7s with no change * ~1.3s with the change, without mmap * 0.04s with the change, with mmap

hghave: add py312 and py313 While not required in the core test suite in the moment, these could be useful in the future or for extensions. For example, Python 3.12 removed distutils and it might make sense to differentiate based on that.

hghave: use strings instead of floats for version numbers passed to checkvers I think it’s a really bad idea to use floats for version numbers. One problem is that 3.10 is the same as 3.1.

py3: fully port doctest to py3

import-checker: make stdlib path detection work in virtual environments The previous logic tried to find the directory containing BaseHTTPServer, which didn’t work as indended because it was only present on Python 2. Instead, the argparse module is used now.

cleanup: remove unnecessary list constructor calls around list comprehensions

branching: merge stable into default

delta-find: pass the full deltainfo to the _DeltaSearch class Having more information is better, so we pass it directly.

delta-find: move sparse-revlog pre-filtering in the associated class Lets move the specialized code in the specialized class.

delta-find: move sparse-revlog delta checks in the associated class Lets move the specialized code in the specialized class.

delta-find: split the _DeltaSearch class in two We now have things sliced small enough to have two class that use different `_iter_groups` implementation to encode their different logic. The filtering code remains to be moved, but I would rather keep this changeset simple and move them in the next.

delta-find: finish reworking the snapshot logic and drop more layer The refining logic only applies to the snapshot logic, and this is now all contained in a dedicated method. Along the way, we drop the refined_groups // raw_groups layer as they no longer make sense. The result is a more explicit `iter_groups` method. This conclude the splitting and simplification of the groups generation. We are now ready to dispatch this in more diverse classes.

delta-find: move the base of the delta search in its own function That logic is complicated enough that is is worth puting in its own function. Another method will be introduced in the next changeset to deal with the actual refining.

delta-find: move the emotion of prev in a dedicated method After splitting the filtering, and with the `_candidate_groups` layer removed, we can start splitting the group generation too. This helps to organize this code and make it easier to modifying the future.

delta-find: move the emotion of parents in a dedicated method After splitting the filtering, and with the `_candidate_groups` layer removed, we can start splitting the group generation too. This helps to organize this code and make it easier to modifying the future.

delta-find: explicitly deal with usage of the cached revision We can remove this from the general logic path and directly deal with this corner case early. This result in a small change in test-generaldelta.t as it turns out that: - at commit time we (sometimes) precompute a delta against p1 and pass it as the cached delta. - since cached delta where going through the same filtering as everything, we could "optimize" the base if it applied to an empty delta, resulting in not using the pre-computed delta. The simpler logic fix the second item, making the cached delta base always actually tested when requested. Note that the computation of a fast delta against p1 only is questionable, but looking into that is out of scope for this series.

delta-find: remove the "candidate groups" layer We have enough pieces to remove this generator and directly bear it load using the underlying object.

delta-find: stop using heuristic to determine if we are creating a snapshot This avoid assuming a changeset is a snapshot when it is actually something simpler.

delta-find: explicitly track stage of the search Being more explicit about what we are doing is going to be useful. We actually start making use of it in later changesets.

delta-find: drop some dead debug code Seems like it was never put to use, so lets simply remove it for now.

delta-find: introduce and use specialized _DeltaSearch class For now, we introduce some very simple variant, but they are still useful to display how having the class can helps keeping the simple case simple and their special case out of more advanced logic.

delta-find: introduce a base class for _DeltaSearch This prepare the introduction of specialized the class in the next changesets.

delta-find: simplify the delta checking function for snapshot Since the function is all about snapshot, we can safely use an early return and make the result simpler.

delta-find: move good delta code earlier in the class Nothing change except the code location. This greatly helps readability of the next future diff,

delta-find: split is_good_delta_info into more thematic function Same logic as for candidate filtering, we group code into related sub method. This will help clarifying later patches as some logic is pre-splitted

delta-find: clarify some comment and code in is_good_delta_info We move the comment closer to the code it describ and we compute an intermediate value without using the `textlen` variable, as it will stop being defined in a future patch. This will clarify future patches.

delta-find: move delta size check earlier in is_good_delta_info This will clarify future patches by regrouping related logic before larger movement.

delta-find: split the delta-chain part of `_pre_filter_rev` in a method Since `_pre_filter_rev` contains logic from various sources of constraint, we start splitting is in subfunction to clarify and document the grouping.

delta-find: split the "sparse" part of `_pre_filter_rev` in a method Since `_pre_filter_rev` contains logic from various sources of constraint, we start splitting is in subfunction to clarify and document the grouping.

delta-find: split the generic part of `_pre_filter_rev` in a method Since `_pre_filter_rev` contains logic from various sources of constraint, we start splitting is in subfunction to clarify and document the grouping.

delta-find: drop the temporary indent Now that the complicated change is made, we can do the noisy one.

delta-find: move pre-filtering of individual revision in its own function This goes one step further than the previous change by making the pre-filtering of individual candicates revision in its own function. This will allow subclass to easily configure this filtering with their own constrains. The `if True:` part help the readability of this diff a lot and will be drop in to the next changesets.

delta-find: move pre-filtering of candidates in its own function This organise the code further and open the way to specialization via sub-classing. Something important for the coming changes.

delta-find: move away from the generator API for _DeltaSearch We use more explicit function call. This make operations more explicit and will make future refactoring simpler.

delta-find: use "-1" as depth snapshot-dept for non snapshot in debug This will help do distinct full snapshot (level 0) and normal delta (not a snapshot, no snapshot level)

delta-find: fix the computation of the `prev` value The previous computation was "wrong" it always used the tiprev, even when computing a delta in a non-append case (mostly benchmark). This never produced wrong delta on disk, but would misled debug or performance command. Since it does not have any actual user impact, I did not put this on stable. With the code fixed we can now use revisions in some search and it makes the test display more interesting behavior since the algorithm has more to work with.

delta-find: move is_good_delta_info on the _DeltaSearch class There is a lot of format specific code in `is_good_delta_info`, moving it on _DeltaSearch will allow to split this into subclass soon.

delta-find: feed revinfo to _DeltaSearch The revinfo has more information and will allow for even more function to be turned into method.

delta-find: clarify that revisioninfo.p1/p2 constains nodeid This clarify the content of these attributes.

delta-find: move filing of some debug data in `_one_dbg_data` Since the `_one_dbg_data` method is meant to create a valid debug dictionnary. We can as well prefill the relevant value to reduce the amount of debug code in the main code.

delta-find: add more explanation to the the deltas_limit < length check More explanations is always good.

delta-find: move tested in the _DeltaSearch.__init__ Now that we have an object we can initialize that attribute at initialization time. This will make it available for more method in the future, allowing to split the code.

delta-find: check DELTA_BASE_REUSE_FORCE in the _DeltaSearch.__init__ Now that we have an object we can check that DELTA_BASE_REUSE_FORCE cases does not reach this code at in a more suitable location.

delta-find: move target_rev in the _DeltaSearch.__init__ Now that we have an object we can initialize that attribute at initialization time.

delta-find: move snapshot_cache in the _DeltaSearch.__init__ Now that we have an object we can initialize that attribute at initialization time.

delta-find: move `_rawgroups` on the `_DeltaSearch` object Moving more code before doing more logic changes.

delta-find: move `_refinedgroups` on the `_DeltaSearch` object Moving more code before doing more logic changes.

delta-find: introduce a _DeltaSearch object That object represent the search of a good delta for one revision. It will replace the interleaved generator currently in use. It will make the logic more explicit and easier to split into different subclass for the algorithm variant. We will move content gradually before doing deeper rework. For now, we only move the `_candidategroups` function here. More will follow in the same series.

delta-find: add a small docstring to deltacomputer As we are about to introduce another object related to finding delta. So lets have a minimal docstring to the existing one.

revlog: stop using `atomictmp` for the split revlog Since we already manually deal with writing on the side and delaying visibily, we no longer need this.

changelog: drop the side_write argument to revlog splitting The only user is now gone.

changelog: stop useless enforcing split at the end of transaction Changelogs are no longer created inline, and existing changelogs are automatically split. Since we now enforce splitting at the start of any write, we don't need to enforce splitting at the end of the transaction. This has the nice side effect of killing the only user of "side_write".

branching: merge stable into default

contrib: add a set of scripts to run pytype in Docker Having a simple way to run pytype for developers can massively shorten development cycle. Using the same Docker image and scripts that we use on our CI guarantees that the result achieved locally will be very similar to (if not the same as) the output of our CI runners. Things to note: the Dockerfile needs to do a little dance around user permissions inside /home/ci-runner/ because: - on one hand, creating new files on the host (e.g. .pyi files inside .pytype/) should use host user's uid and gid - on the other hand, when we run the image as uid:gid of host user, it needs to be able to read/execute files inside the image that are owned by ci-runner Since local user's uid might be different from ci-runner's uid, we execute this very broad chmod command inside /home/ci-runner/, but then run the image as the host user's uid:gid. There might be a better way to do this.

pytype: use "$(hg root)" instead of `hg root` to make shellcheck happier

pytype: update check-pytype.sh to select target automatically We have python3.11 on CI, so we can run pytype targeting that version. On the other hand, we don't have python3.7 on CI anymore, so we can't run pytype for 3.7 anymore (interpreter not found). I think it's fine to make pytype select the appropriate target depending on the version of the interpreter it's running under.

git-hgext: adjust to the lack of `changelog.heads` method We don't have a `heads` method returning nodeid, but this is very easy to get the same result. This was flagged by pytype. We can note that the fact this code did not break is probably a good sign that it is dead code. However this is a question outside of the scop of this series.

remotefilelog: drop dead code As pytype flagged bug in this method it highlighted that this methode being never called anywhere.

pytype: use the right signature for the `__delitem__` It is not because it is NotImplemented that it should use a bad signature. Fix it to please pytype.

pytype: use the right signature for the `__setitem__` It is not because it is NotImplemented that it should use a bad signature. Fix it to please pytype.