store: use the boolean property in `store`

store: introduce boolean property for revlog type This will avoid exposing implementation details to more generic code.

store: issue a single entry for each revlog We now yield a single entry, with data about each files involved. This help to simplify multiple code using this and it will help to further simplify and fixes the streaming code.

store: rename `unencoded_path` to `entry_path` for StoreEntry This remove the ambiguity with StoreFile and make sure use code will be using the right API.

store: do the revlog matching on entry directly This is the last blocker to safely merge the revlog files in a single entry.

store: split the wrapping of encodedstore between _wrap and datafiles The `datafiles` method of `basicstore` is doing a lot of work that should be done on decoded filename. So we now wrap `_walk` to do the decoding, and less work in `datafiles`. This is necessary to make sure file from the same revlog can be grouped together.

store: introduce a main_file_path method for revlog This help code that need to point revlog to an index file. This is put to use in the upgrade code.

upgrade: actually use StoreEntry API to create revlog Lets make use of the semanctic of the object we are passed.

upgrade: use StoreEntry object in upgrade We will make more use of the API in the next changeset, but just moving to use entry is a significant change for the engine codebase.

upgrade: drop a quick fix that is not longer necessary We won't issue bad revlog from topfile anymore.

store: use StoreEntry API instead of parsing filename in largefile This is more explicit and more robust.

store: use StoreEntry API instead of parsing filename when listing manifestlog This is more explicit and more robust.

store: use StoreEntry API instead of parsing filename when fixing issue6528 This is more explicit and more robust. We also introduce a small output change as it make things simpler and this is a affecting a debug-command.

store: use StoreEntry API instead of parsing filename in remotefilelog This is more explicit and more robust.

store: use StoreEntry API instead of parsing filename in narrow This is more explicit and more robust.

store: add a `target_id` attribute on RevlogStoreEntry This hold the "target" (file, directory, etc) of a revlog. Having this available will help a lot of code to avoid direct file path access.

store: actually tag tree manifest revlogs as manifest revlogs It turn out we have been mislabeling these for a long while. This is now fixed.

store: also gather files per revlog in `topfiles` This conclude out revlog gathering.

store: also group files by revlog in fncache version of datafiles One more step.

store: add logic to group revlog file together For now each file get its own entry, this will help stopping this, soon™. We use such gathering in the `basicstore` code.

store: change `_walk` return to `(filename, (type, size))` If we are to group file per revlog, having the filename as the "main key" will be useful. This change will make the following changes clearer.

store: lazily get file size on demand for the fncache case We don't have the information in the first place, so we can avoid querying the file system inconditionnaly for use case we don't needs it. This change requires the StoreFile class to be passed a vfs to retrieve the file_size if needed. In the non-fncache case, we already have the information from file system walking, so we keep it and use it.

store: only access is_volatile information through the file object This make sure other code only access this information through the proper API, and it prepare for store entries to be able to agregate multiple files.

store: only access file_size information through the file object This make sure other code only access this information through the proper API, and it prepare for store entries to be able to agregate multiple files.

store: have custom init for entries class This will get useful to add special processing later in this series.

store: use specialized class for store entries We introduce two different classes for revlog and other entries. For now, we still have multiple entry for the same revlog, but we will work toward grouping the different file in a single entry in this series. Having the distinction is a step toward this goal.

store: introduce a EntryFile object to actually access file info For now a StoreEntry match a single file, but the goal is to eventually combine multiple file in a higher level Entry, so we need to introduce this distinction and use it first.

store: use a StoreEntry object instead of tuple for store files We want to make the store return more semantic information instead of a stream of file path. To achieve this, we start with adding a simple object that hold the same information as the tuple it replace, and do a simple update to the user code to fetch and use the same information. From there, we will be able to iteratively upgrade the codebase toward better objects.

store: no longer explicitly mark the splitted index as ignored Since the fncache is going to automatically recognised it as non "normal" revlog file, we don't need to explicitly make it ignored.

store: cleanup what is recognized as a revlog file The temporary file from `censor` are not relevant to the store as they are just temporary. So let `fncache` and `datafiles` forget about them.

store: only inclure revlog related file in `datafiles` This is a "reading" equivalent of the previous patches (see that patch for details).

store: do not record file that are not part of a revlog in fncache The fncache exist to list file to copy/stream when cloning a repository, it should only contains file that are relevant for a revlog in such case. For example, temporary file are not relevant. So we now skip the addiction of non-relevant file to the fn-cache in the first place.

revlog-split: expand the some test to cover the success case too It seems useful to double check the case of a transaction successfully committed before trying the one with transaction abort.

transaction: use a ".bck" extension for all backup file There are too much code in Mercurial that assume file function from extension. Keeping the original extension when doing backup is just creating confusion (for exemple, backup of revlog's confused with new version revlogs). So we change the backup name from now on to be explicilty backup file and prevent this kind of problem. (note that repository using fncache is less affected (but still affected) by this kind of problem as the backup are not listed in the fncache).

repair-issue6528: remove a now useless continue The condition for that continue have been excplicly filtered out of the `files` list a handful of line earlier.

rebase: don't call rewriteutil.precheck() with to-be-skipped commits It's clearly incorrect to call `rewriteutil.precheck()` for commits that we're not about to rewrite. We haven't noticed yet because the function doesn't check for divergence, but I'm about to teach it to do that. Differential Revision: https://phab.mercurial-scm.org/D10259

copies: make sure deleted copy info do not overwriting unrelated ones See test documentation for details. This is yet another corner case for copy tracing. Differential Revision: https://phab.mercurial-scm.org/D9200

utils: helper function to print top memory allocation site The memorytop function uses Python's tracemalloc module to show the source lines / backtraces with the largest remaining allocations. This allows identifying the origins of active memory by placing calls in strategic locations. Allocations from C extensions will show up as long as they are using the Python allocators. Differential Revision: https://phab.mercurial-scm.org/D9236

branching: merge stable into default

stabletailgraph: clarify omission of linear parts in test sorts Those are not interesting and are collapsed using globs. Only the tagged nodes are interesting.

stabletailgraph: omit uninteresting linear parts in test log The test uses tags for the nodes of interest. There is no need to print the linear parts which exist only to force the rank.

stabletailgraph: fix typo in test

stabletailgraph: fix terminology in doc

Added signature for changeset fc445f8abcf9

Added tag 6.4.3 for changeset fc445f8abcf9

relnotes: add 6.4.3

backup: fix issue when the backup end up in a different directory Because of store encoding, we might end up with the backup in a different directory than the initial copy (for example if the backup path make it cross the 120 char limit). This can create crash, especially since 6.4 where backup are used during revlog split. Making sure the directory exists fixes these crash We added a test covering this case. Strictly speaking, this has always been broken, however the new code in 6.4 triggers it more easily.

vfsproxy: inherit the `createmode` attribute too It is an important part of the API when creating directory. We will need it in the next changeset.

revlog: test more complex file pattern for revlog split There have been a report of failure while splitting revlog. The file name involved was important. The pattern involved here are not problematic, but it help to setup the machinery to test multiple files, the actual fix and problematic file will arrive in a later changeset.

repo-upgrade: write new requirement before upgrading the dirstate This will prevent a small race condition where another hg process still believes the repo is dirstate-v1 during the upgrade process. This is good to have, but it is not a proper fix for the underlying problem. There is code that assumes a requirement means a usage, e.g. having the `generaldelta` requirement would imply *all* revlogs to use general delta, but it's not true, it simply means that the repository advertises to the client it needs to understand `generaldelta` in order to read the repo. In the case of the dirstate, having the requirement *technically* should always be the same as using dirstate-v2, since there is only one dirstate and requirements should be as minimal as possible. However, we should not assume this and make the code more robust in a future patch (series).

perf: introduce a `perf::stream-locked-section` command This command benchmark the initial part of a stream clone, where the repository is locked.

transaction: tr._names are actually bytes, use byte string to join them Looks like this is never checked in the tests? But trying e.g. `print(tr)` by hand before this change would give: TypeError: sequence item 0: expected str instance, bytes found Looks like tr._names are always bytes (although this isn't actually enforced, but maybe at some point it will at least be type checked e.g. by pytype).

nix: add nix-defined package and devel env (flake) This adds a Nix Flake which defines: - package recipies to build and run Mercurial (C, C+Rust variants), - a meta package to run the test suite in an isolated sandbox, - a development environment with some tools (notably Python and the Black formatter, pinned to match the CI). Python is pinned to the recommended version. The Rust toolchain is a fairly recent one provided by the Nixpkgs channel. It is not yet pinned to the same version as the "reference" Debian package, but this does not seem to cause any issue. Example usage of local commands are provided in the `flake.nix` file. Once merged, it should also be possible to pull and run directly from the source repository directly with something like: `nix run hg+https://foss.heptapod.net/mercurial/mercurial-devel?ref=TOPIC&dir=contrib/nix' -- version`

hgignore: ignore nix result symlink

tests: add test ignore list for nix Some tests cannot run properly in the nix sandbox. This adds a list of tests to ignore when running in that context.

ui: keep the progress bar around when writing if stdout is not a tty

rhg: correctly relativize copy source path

rhg: don't print copy source when --no-status is passed

tests: add test for status copy source formatting

fix: highlight the required configuration and behavior of the fixer tools The problem is that `hg help fix` didn't mention *how* to configure the tools, and while I knew that `{rootpath}` existed in the configuration, I missed that the tools require reading content from stdin. (I configured `gofmt` to use `{rootpath}`, and that had the effect of squashing all changes in a file at `.` into the first commit and emptying that content from its descendants.) Basically all this does is put a pointer in the default (command level) help to the extension level help that mentions the configuration, and moves the extension level help that documents reading from stdin and writing to stdout to the top to give it more prominence. The last sentence is adjusted a bit to reflect the new location.

debug-delta: add minimal documentation for `devel.bundle-delta` option So the next person won't have to grep the code to know what it is doing.

delta-find: simply code that is now never invoqued without general delta Finish early in on-general delta case mean a lot of code can assume that general delta is not enabled.

delta-find: add a simple safeguard to prevent bad non-general-delta This make it easier to catch request to build delta that does not make sense in the general delta world. It seems better to raise a programming error than to corrupt the repository.

delta-find: never do anything fancy when general delta is off Without general delta, the only option is to do a delta against the previous revision in the revlog, or to do a full snapshot. So no need to dive into the more complicated logic. This prevent Mercurial to misbehave when applying remote deltas when `pulled-delta-reuse-policy` was set to "try-base" or "forced" and would produce and store invalid deltas. We add an associated tests to make sure this scenario is covered by tests. There are more code to cleanup after this, but I what to keep the fixing patch simple.

delta-find: assume the target-rev if not specified This will be useful in the next patch, but I am putting it in its own commit to make sure we can easily bisect any subtle side effect. (none are expected, but still)

tests: in filterpyflakes, tolerate non-ascii file contents

comments: fix spelling

chg: set CHGHG before connecting to command server cf4d2f31 (!523) changed chg to set `CHGHG` itself when spawning a new command server, in order to ensure that the path to the `hg` executable would be checked during server validation. (This is useful when chg is built with `HGPATHREL`). However, that change broke chg because it failed to set `CHGHG` before trying to connect to an existing command server. This means that if `CHGHG` is not present in the environment, chg will always spawn a new command server, entirely negating the point of chg. This breakage wasn't initially caught because of the difficulty of writing automated tests with the `HGPATHREL` feature enabled, which meant the change was only tested manually to make sure that it fixed the problem with `HGPATHREL` that prompted the change. In practice, this functionality is only really useful when chg is built with `HGPATHREL`, so I considered wrapping it in an `#ifdef` to preserve the old behavior by default. However, this makes it hard to write tests since one would have to explicitly set `HGPATHREL=1` when running `run-tests.py` (which is why the original change lacked tests). It would be great if there were a way of testing features that are gated behind conditional compilation.

chg: declare environ (issue6812)

heptapod-ci: display python version it will be useful to debug in the CI

branching: merge stable into default

relnotes: add 6.4.2

Added signature for changeset f952be90b051

Added tag 6.4.2 for changeset f952be90b051

revlog-split: make sure the self._indexfile attribut is reset (issue6811) Before this change, after a transaction committing a file split, a revlog object would have its `self._indexfile` attribute desynchronised from the actual file storing the data. If that same object is reused (as we do for the manifest during clone bundles), this lead to the data being writting in the wrong location and the repository to go corrupt. We not properly reset the attribut when applicable and everything is back in working order.

revlog-split: also check a case where bundle contains multiple changegroup This case is not affected, we remains in the same transaction. However it seems important to improve the test coverage around these "features". It also helped to assert the scop of the critical bug.

revlog-split: show manifest data loss situation when using clonebundle After the commit of the transaction from applying the clone bundle, changes to the manifest from the subsequent pull would write index data in the wrong file. This lead to data being "lost" after the pull and the repository to be corrupted. We add a test checking this specific scenario. In practice, the issue only affects the manifest but we also test the issue with a file for future proofing things up. See test documentation and failure for details.

sha1dc: Make sure SHA1DC_BIGENDIAN is set on Darwin/PowerPC __BYTE_ORDER__ is not defined on old toolchains so sha1dc gets built in little endian mode on a big endian platform, hilarity ensues. Tested on Mac OS X Tiger 10.4.

hidden: add support to explicitly access hidden changesets with SSH peers This implements support for using --remote-hidden with an SSH server. The remote `hg serve --stdio` call is passed the `--hidden` flag as a request to access hidden changesets. This approach has benefits similar to the one we used for HTTP peers. It * works around the lack of global parameters in wire protocol v1, * reuses the `--hidden` flag (that does not use the wireproto), and * can be safely ignored by older client (fitting the best effort contract). Same as for HTTP, the feature is experimental so we have all the room we needs to update the implementation in the future if deemed necessary. The SSH version of the `--remote-hidden` config uses the same configuration as the HTTP support to control the access to this feature. The name of the user running the command is used for the checking. Test written by Pierre-Yves David.

hidden: support passing --hidden with `serve --stdio` This currently has no effect since the server code will filter the repository again. However, it is clearer to put this change in its own changeset. This will mostly impact ssh peers, clients being now able to pass --hidden when spawning the server. See the next changeset for details on how this is going to be used.

hgweb: move ismember from `hgweb.common` to `scmutil` To implement hidden access control for other protocol than http, we need code outside of hgweb to access this logic.

hidden: add support for --remote-hidden to HTTP peer Test written by Pierre-Yves David.

hidden: add support to explicitly access hidden changesets via HTTP We now detect hidden access request and validate them while handling HTTP wireprotocol command. This changeset only introduces server support. Client support is added in a subsequent changeset. The mechanism we use for the HTTP wireprotocol is the same as the one used to lift filtering for hgweb. The use of the global `access-hidden` parameter in the http url is a simple and effective approach that * works around the lack of global parameters in wire protocol v1, * reuses the mechanism used for hgweb (that does not use the wireproto), and * can be safely ignored by older client (fitting the best effort contract). The feature is still experimental, so we can easily update the implementation details in the future. However, landing a first implementation now will help test the feature in the wild and iterate on alternative implementations if deemed necessary.

hidden: support for explicitly accessing hidden changesets in wireproto server This installs the basic infrastructure to be able to access hidden changeset through the wireprotocol. Each wireprotocol server still needs a way to define how the feature is triggered.

hidden: add support for --remote-hidden to localpeer

pull: add --remote-hidden option and pass it through peer creation This option will allow to pull changesets that are hidden on the remote. This is useful when looking into a changeset’s evolution history, resolving evolution instability or mirroring a repository. The option is best effort and will only affect the pull when it can. The option will be ignored when it cannot be honored. Support for each type of peer is yet to be implemented. They currently all warn about lack of support. The warning code will get removed as peers gain support for this option. The option is still experimental, so we will have freedom to update the UI or implementation before it graduates out of experimental. Based on a changeset by Pierre-Yves David, which added the option.

hgweb: add support to explicitly access hidden changesets This changeset adds a "global" `access-hidden` argument to hgweb. This argument lift the "hidden" filtering. This means the request has access to hidden (eg: obsolete) changesets. Secret changesets remains filtered. This feature has multiple applications. The first main use case is to allow the hgweb interface to display more obsolescence related data, such as the Anton Shestakov work to add `obslog` support to hgweb. The second foreseen usecase is support for a `--remote-hidden` argument to `hg pull` and `hg clone`. This flag will make it possible to retrieve hidden (typically obsolete) changeset under some conditions. This is useful when digging up obsolescence history or when doing full mirroring. More on this feature coming in later changesets. To avoid exposing information by mistake, access to this feature is currently controlled with the `experimental.server.allow-hidden-access` config option. The option works the same way as `web.allow-push`. The current default is to not allow any hidden access. However we might change it before the feature stop being experimental.

clone-bundles: add an option to generate bundles in the background This is what the "on-change" version have been doing from the start, it seems useful to also offer this option in the command.

clone-bundles: add a command to clear all bundles

clone-bundles: add more test cases for automatic generation See inline documentation for details.

clone-bundles: introduce a command to refresh bundle This command allows for a different mode than the one attached to the transation. It might fit some hosting setup better.

clone-bundles: add a configuration to control auto-generation on changes We are about to introduce a "manual" way to deal with automatic clone management but running a command using some internal API. The first step is to introduce a way to control the "on changes" behavior.

clone-bundles: also control automation based on absolute number of revisions

clone-bundles: only regenerate the clone bundle when cached ration is low See inline documentation for details.

clone-bundles: garbage collect older bundle when generating new ones See inline documentation for details.

clone-bundles: add a basic first version of automatic bundle generation See inline documentation for details.

branching: merge stable into default

Added signature for changeset 83ea6ce48b4f

Added tag 6.4.1 for changeset 83ea6ce48b4f

relnotes: add 6.4.1

sslutil: set context security level for legacy tls testing (issue6760) Current versions of OpenSSL do not allow the use of TLS <1.2 when the library's security level is >=1 (1 being the default on most distributions). Setting the security level in addition to the minimum protocol is therefore necessary for the legacy protocol tests. This is done here ONLY when testing, when: - explicitly setting the cipher string, or - using the "--insecure" flag, or - using the "devel.serverexactprotocol" testing option. See: https://github.com/openssl/openssl/blob/master/NEWS.md#major-changes-between-openssl-30-and-openssl-310-14-mar-2023

configitems: make devel.serverexactprotocol look dangerous Because it is.

rebase: do not cleanup the working copy when --dry-run is used (issue6802) Since we did not touch the working copy, we don't need to clean it up. This will avoid wiping exiting changes out.

rebase: add a test showing that --dry-run wipes working copy changes Eating people's data on --dry-run seems like a bad idea.

tests: automatically glob the discovery timing information Time is not stable in tests.

rhg: support `status --print0` This seems very easy to support, and useful because it makes it possible to parse the [hg status] output even if the user creates files with '\n' characters by accident.

stabletailgraph: implement stable-tail sort This adds the computation of the "stable-tail sort", an incremental node sorting method. It is a stepping stone for the implementation of faster label discovery (for example for obs markers) and more caching.

heptapod: add `.gitattributes` file to improve language detection I am fully aware of the irony.

match: match explicit file using a set The matcher as all the logic to do quick comparison against explicit patterns, however the pattern matcher was shadowing the code using that set and used the compiled regex pattern in all cases, which is quite slow. We restore the usage of the set based matching to boost performance. Building the regexp is still consuming a large amount of time (actually, the majority of the time), which is still silly. Maybe using re2 would help that, but this is a quest for another adventure. Another path to improve this is to have a pattern type dedicated to match the exact path to a file only (not a directory). This pattern could use the set matching only and be skipped in the regex all together. Benchmarks ========== In the following benchmark we are comparing the `hg cat` and `hg files` run time when matching against all files in the repository. They are run: - without the rust extensions - with the standard python engine (so without re2) Performance improvement in this series -------------------------------------- ###### hg files ############################################################### ### mercurial-2018-08-01-zstd-sparse-revlog ### sorted base-changeset: 0.230092 seconds prev-changeset: 0.230069 seconds this-changeset: 0.211425 seconds (-8.36%) ### mercurial-2018-08-01-zstd-sparse-revlog ### shuffled base-changeset: 0.234235 seconds prev-changeset: 0.231165 seconds (-1.38%) this-changeset: 0.212300 seconds (-9.43%) ### pypy-2018-08-01-zstd-sparse-revlog ### sorted base-changeset: 0.613567 seconds prev-changeset: 0.616799 seconds this-changeset: 0.510852 seconds (-16.82%) ### pypy-2018-08-01-zstd-sparse-revlog ### shuffled base-changeset: 0.801880 seconds prev-changeset: 0.616393 seconds (-23.22%) this-changeset: 0.511903 seconds (-36.23%) ### netbeans-2018-08-01-zstd-sparse-revlog ### sorted base-changeset: 21.541828 seconds prev-changeset: 21.586773 seconds this-changeset: 13.648347 seconds (-36.76%) ### netbeans-2018-08-01-zstd-sparse-revlog ### shuffled base-changeset: 172.759857 seconds prev-changeset: 21.908197 seconds (-87.32%) this-changeset: 13.945110 seconds (-91.93%) ### mozilla-central-2018-08-01-zstd-sparse-revlog ### sorted base-changeset: 62.474221 seconds prev-changeset: 61.279490 seconds (-1.22%) this-changeset: 29.529469 seconds (-52.40%) ### mozilla-central-2018-08-01-zstd-sparse-revlog ### shuffled base-changeset: 1364.180218 seconds prev-changeset: 62.473549 seconds (-95.40%) this-changeset: 30.625249 seconds (-97.75%) ###### hg cat ################################################################# ### mercurial-2018-08-01-zstd-sparse-revlog ### sorted base-changeset: 0.764407 seconds prev-changeset: 0.763883 seconds this-changeset: 0.737326 seconds (-3.68%) ### mercurial-2018-08-01-zstd-sparse-revlog ### shuffled base-changeset: 0.768924 seconds prev-changeset: 0.765848 seconds this-changeset: 0.174d0b seconds (-4.44%) ### pypy-2018-08-01-zstd-sparse-revlog ### sorted base-changeset: 2.065220 seconds prev-changeset: 2.070498 seconds this-changeset: 1.939482 seconds (-6.08%) ### pypy-2018-08-01-zstd-sparse-revlog ### shuffled base-changeset: 2.276388 seconds prev-changeset: 2.069197 seconds (-9.15%) this-changeset: 1.931746 seconds (-15.19%) ### netbeans-2018-08-01-zstd-sparse-revlog ### sorted base-changeset: 40.967983 seconds prev-changeset: 41.392423 seconds this-changeset: 32.181681 seconds (-22.20%) ### netbeans-2018-08-01-zstd-sparse-revlog ### shuffled base-changeset: 216.388709 seconds prev-changeset: 41.648689 seconds (-80.88%) this-changeset: 32.580817 seconds (-85.04%) ### mozilla-central-2018-08-01-zstd-sparse-revlog ### sorted base-changeset: 105.228510 seconds prev-changeset: 103.315670 seconds (-1.23%) this-changeset: 69.416118 seconds (-33.64%) ### mozilla-central-2018-08-01-zstd-sparse-revlog ### shuffled base-changeset: 1448.722784 seconds prev-changeset: 104.369358 seconds (-92.80%) this-changeset: 70.554789 seconds (-95.13%) Different way to list the same data with this revision ------------------------------------------------------ ###### hg files ############################################################### ### mercurial-2018-08-01-zstd-sparse-revlog root: 0.119182 seconds glob: 0.120697 seconds (+1.27%) sorted: 0.211425 seconds (+77.40%) shuffled: 0.212300 seconds (+78.13%) ### pypy-2018-08-01-zstd-sparse-revlog root: 0.121986 seconds glob: 0.124822 seconds (+2.32%) sorted: 0.510852 seconds (+318.78%) shuffled: 0.511903 seconds (+319.64%) ### netbeans-2018-08-01-zstd-sparse-revlog root: 0.173984 seconds glob: 0.227203 seconds (+30.59%) sorted: 13.648347 seconds (+7744.59%) shuffled: 13.945110 seconds (+7915.16%) ### mozilla-central-2018-08-01-zstd-sparse-revlog root: 0.366463 seconds glob: 0.491030 seconds (+33.99%) sorted: 29.529469 seconds (+7957.96%) shuffled: 30.625249 seconds (+8256.97%) ###### hg cat ################################################################# ### mercurial-2018-08-01-zstd-sparse-revlog glob: 0.647471 seconds root: 0.643120 seconds shuffled: 0.174d0b seconds (+13.92%) sorted: 0.737326 seconds (+13.88%) ### mozilla-central-2018-08-01-zstd-sparse-revlog glob: 40.596983 seconds root: 40.129136 seconds shuffled: 70.554789 seconds (+73.79%) sorted: 69.416118 seconds (+70.99%) ### netbeans-2018-08-01-zstd-sparse-revlog glob: 18.777924 seconds root: 18.613905 seconds shuffled: 32.580817 seconds (+73.51%) sorted: 32.181681 seconds (+71.38%) ### pypy-2018-08-01-zstd-sparse-revlog glob: 1.555319 seconds root: 1.536534 seconds shuffled: 1.931746 seconds (+24.20%) sorted: 1.939482 seconds (+24.70%)

match: sort patterns before compiling them into a regex While investigating cripping performance for `hg cat` in some context, I discovered that, for large inputs, building a regex from out of order patterns result may result in a *much* slower regex and a much slower associated matcher's performance. So we are now sorting the patterns to help the regex engine. There is more to the story as we rely on regexp more than we should. See the next changeset for details. Benchmarks ========== In the following benchmark we are comparing the `hg cat` and `hg files` run time when matching against the full list of files in the repository. They are run: - without the rust extensions - with the standard python enfine (so without re2) sort vs non-sorted - Before this changeset (3f5137543773) --------------------------------------------------------- ###### hg files ############################################################### ### mercurial-2018-08-01-zstd-sparse-revlog sorted: 0.230092 seconds shuffled: 0.234235 seconds (+1.80%) ### pypy-2018-08-01-zstd-sparse-revlog sorted: 0.613567 seconds shuffled: 0.801880 seconds (+30.69%) ### mozilla-central-2018-08-01-zstd-sparse-revlog sorted: 62.474221 seconds shuffled: 1364.180218 seconds (+2083.59%) ### netbeans-2018-08-01-zstd-sparse-revlog sorted: 21.541828 seconds shuffled: 172.759857 seconds (+701.97%) ###### hg cat ################################################################# ### mercurial-2018-08-01-zstd-sparse-revlog sorted: 0.764407 seconds shuffled: 0.768924 seconds ### pypy-2018-08-01-zstd-sparse-revlog sorted: 2.065220 seconds shuffled: 2.276388 seconds (+10.22%) ### netbeans-2018-08-01-zstd-sparse-revlog sorted: 40.967983 seconds shuffled: 216.388709 seconds (+428.19%) ### mozilla-central-2018-08-01-zstd-sparse-revlog sorted: 105.228510 seconds shuffled: 1448.722784 seconds (+1276.74%) sort vs non-sorted - With this changeset ---------------------------------------- ###### hg files ############################################################### ### mercurial-2018-08-01-zstd-sparse-revlog all-list-pattern-sorted: 0.230069 all-list-pattern-shuffled: 0.231165 ### pypy-2018-08-01-zstd-sparse-revlog all-list-pattern-sorted: 0.616799 all-list-pattern-shuffled: 0.616393 ### netbeans-2018-08-01-zstd-sparse-revlog all-list-pattern-sorted: 21.586773 all-list-pattern-shuffled: 21.908197 ### mozilla-central-2018-08-01-zstd-sparse-revlog all-list-pattern-sorted: 61.279490 all-list-pattern-shuffled: 62.473549 ###### hg cat ################################################################# ### mercurial-2018-08-01-zstd-sparse-revlog sorted: 0.763883 seconds shuffled: 0.765848 seconds ### pypy-2018-08-01-zstd-sparse-revlog sorted: 2.070498 seconds shuffled: 2.069197 seconds ### netbeans-2018-08-01-zstd-sparse-revlog sorted: 41.392423 seconds shuffled: 41.648689 seconds ### mozilla-central-2018-08-01-zstd-sparse-revlog sorted: 103.315670 seconds shuffled: 104.369358 seconds

peer: rename makepeer() → _make_peer() In httppeer and sshpeer, there previously were makepeer() and make_peer(), which was confusing. Therefore, this changeset renames one of the functions. makepeer() was the internal function called by make_peer() and some debug command. This function is renamed to _make_peer().

rust: configure MSRV in Clippy This setting makes Clippy never apply lints that are meant for later versions. In case the target precise toolchain is the one running, it does not make a difference, but this gives us a machine-parseable specification that is pretty standard. The README and `hg help rust` are updated to state that `clippy.toml` is the single source of truth about that, also lifting a minor ambiguity: it is fine if the MSRV is lagging behind the version in Debian testing.

rust-readme: mentioned that format check is enforced by CI

rust-readme: mentioning clippy especially since there is a CI check for it.

rust-readme: rst fixes

chg: populate CHGHG if not set Normally, chg determines which `hg` executable to use by first consulting the `$CHGHG` and `$HG` environment variables, and if neither are present defaults to the `hg` found in the user's `$PATH`. If built with the `HGPATHREL` compiler flag, chg will instead assume that there exists an `hg` executable in the same directory as the `chg` binary and attempt to use that. This can cause problems in situations where there are multiple actively-used Mercurial installations on the same system. When a `chg` client connects to a running command server, the server process performs some basic validation to determine whether a new command server needs to be spawned. These checks include things like checking certain "sensitive" environment variables and config sections, as well as checking whether the mtime of the extensions, hg's `__version__.py` module, and the Python interpreter have changed. Crucially, the command server doesn't explicitly check whether the executable it is running from matches the executable that the `chg` client would have otherwise invoked had there been no existing command server process. Without `HGPATHREL`, this still gets implicitly checked during the validation step, because the only way to specify an alternate hg executable (apart from `$PATH`) is via the `$CHGHG` and `$HG` environment variables, both of which are checked. With `HGPATHREL`, however, the command server has no way of knowing which hg executable the client would have run. This means that a client located at `/version_B/bin/chg` will happily connect to a command server running `/version_A/bin/hg` instead of `/version_B/bin/hg` as expected. A simple solution is to have the client set `$CHGHG` itself, which then allows the command server's environment validation to work as intended. I have tested this manually using two locally built hg installations and it seems to work with no ill effects. That said, I'm not sure how to write an automated test for this since the `chg` available to the tests isn't even built with the `HGPATHREL` compiler flag to begin with.

run-tests: remove obsolete coverage check and packaging import (issue6805) This removes an obsolete `coverage` version check (version from a decade ago). This also conveniently removes the dependency over `packaging.version`, which requires some additional installation since Python 3.10.

test-tx-rollback: more lenient glob for kill status (issue6807) The "killed" message may have some prefix and/or suffix which differ depending on the platform. This makes the pinned test output more lenient to accept those.

rustdoc: nodemap doc refreshing Not pretending to be comprehensive. - correcting some inconsistencies - adding a few missing doc-comments - adding more cross references (in some cases it's right beside the current documentation item, but it will nevertheless also be useful, because `rustdoc` will warn us if inconsistencies arise).

rustdoc: fixed or introduced crossrefs in nodemap.rs

rustdoc: summary line for hg_path_to_os_string The main motivation of this change is to avoid the TODO being the summary line, even though this leads to a pretty obvious summary. Then doc-comments for the other functions are introduced for consistency.

rustdoc: wording for checkexec Notably separating the summary line for correct display at module level.

rustdoc: fixed warnings about links This is the minimal fix making those that actually were supposed to be links to work (including in private items).

rust-changelog: introduce ChangelogEntry parent entries accessors Straightforwards now that lifetimes are explicit in `RevlogEntry` parent accessors.

rust-revlog: fix lifetime problem for RevlogEntry parent entries accessors Without this, the lifetime of the result is equated to the lifetime of the `self` reference, preventing callers, e.g., to take a `RevlogEntry` and return its `p1_entry()`, as it looks like returning something that does not outlive the *reference to* the `RevlogEntry`.

rust-revlog: explicit naming for `RevlogEntry` lifetime This matches what has been done in `revlog::changelog::ChangelogRevisionData`, and has the advantage of making things clearer when we introduce other, shorter lived lifetimes.

rust-changelog: introducing an intermediate `ChangelogEntry` Before this change, client code needing to extract, e.g, the Node ID and the description from a changeset had no other choice than calling both `entry_for_rev()` and `data_for_rev()`. This duplicates some (limited) computation, and more importantly imposes bad hygiene for client code: at some point of developement, the client code would have to pass over both entry and data in its internal layers, which at some point of development would raise the question whether they are consistent. We introduce the intermediate `ChangelogEntry` from which both conversion to the generic `RevlogEntry` and extraction of `ChangelogRevisionData` are possible. It might grow some convenience methods in the future. We keep the `data_for_rev()` method of `Changelog` for compatibility, pointing users at the more powerful alternative.

rust-changelog: added a test for `NULL_REVISION` special case The result is due to `Revlog.get_rev_data()` returning an empty byte string for `NULL_REVISION`, followed by special case for emtpty byte strings in `ChangelogRevisionData::new()`.

rust-changelog: made doc-comments more consistent The most important is the one about `data_for_rev`, that looked like a copy-paste leftover (got me confused first time I read this code, before I actually learned there were both `Entry` and RevisionData`. In the comment for the `struct`, "changelog" was probably more about the format in general (as documented elsewhere) than as an identifier. Some of the "Return something" had "of", half had "for".

commands: correct documentation of hg serve’s --ipv6 option When the --ipv6 option is given, the server doesn’t listen to a IPv4 socket. This can be verified by running two servers, one with and one without the option, which works fine. I think that listening to both a IPv4 and a IPv6 socket would be better, but given that the Python standard library class underlying the HTTP server supports only one socket, this is not trivial.

rhg: don't crash on empty directory names in path_encode, just in case I don't expect that to be possible, but there's nothing in path_encode.rs that prevents it, and the old code didn't crash in this case, so it's better to be defensive.

rhg: fix a bug in path encoding, demonstrated in the parent commit

rhg: show a bug in the rust implementation of path_encode introduced recently In commit 96d31efd21f7 I did a refactoring where I dropped a chunk of code by accident, thus introducing a bug. This commit adds a test demonstrating that bug.

extras: re-use Projection from jaraco.collections

extras: expose 'retained_extras' for extensions to extend

extras: extract function for projecting names from a gettable object

typing: correct the signature of error.CommandError There's a place in `mercurial.dispatch._parse()` that passes None if a parse error happens before the command can be parsed out, and casting the error to bytes works fine because the command and message fields are apparently ignored. Likewise, TortoiseHg similarly passes None for the same reason.

hgweb: add "children" into the JSON template for a changeset The children are shown in the HTML templates, so I don’t see a reason against including them in the JSON template as well.

branching: merge stable into default

Added signature for changeset f14864fffdca

Added tag 6.4 for changeset f14864fffdca

relnotes: do 6.4

hooks: invalidate the repo after the hooks Since the hooks may have changed the repository content it seems safer to invalidate it. The invalidation is "soft", the data are kept around and few work will be needed to restore them if nothing actually changed.

dirstate: try refreshing the changelog when parent are unknown See inline comment for details.

localrepo: add a `currentlock` method It mirrors he `currentwlock` function.

dirstate: add a test to highlight another changelog / dirstate race

rust: fix thread cap (for real this time) Both e2f8ed37201c and c52435820bbd failed to put a *default* ceiling on the number of threads used by Rayon to prevent a contention issue. Calling `rayon::available_parallelism()` creates the global threadpool, which made our whole dance useless last time.

help: fix formatting for http config section

tests: accept a test output change in [tests/test-serve.t] This fixes a breakage introduced in adecb1ab4a0d. It was not caught by the CI probably because allows binding to port 13.

py3: fix for Python 3.12 emitting SyntaxWarning on invalid escape sequences Missed in 805d4a462abb: $ python3.12 mercurial/store.py mercurial/store.py:406: SyntaxWarning: invalid escape sequence '\.' EXCLUDED = re.compile(b'.*undo\.[^/]+\.(nd?|i)$')

url: don't ignore timeout for https connections For http, we use the stdlib's HTTPConnection.connect which passes the timeout down to socket.create_connection; for https, we override the connect method but weren't handling the timeout, so connections could hang for hours even with http.timeout set to low values.

debugdeltachain: stop summing the same chain over and over Before this patch, delta chain size was computed from scratch for each chain, disregarding the fact very likely already computed the same of length-1 prefix for another revisions. We not cache delta chain size and shortcut the computation when we see them. Just for my mercurial-devel clone, this move the computation from about 17.5 second to about 4.8 seconds.

revlog: improve the robustness of the splitting process The previous "in-place" splitting, preserving the splitting on transaction failure had a couple of issue in case of transaction rollback: - a race windows that could still lead to a crash and data loss - it corrupted the `fncache`. So instead, we use a new approach that we summarized as "we do a backup of the inline revlog pre-split, and we restore this in case of failure". To make readers live easier, we don't overwrite the inline index file until transaction finalization. (once the transaction get into its finalization phase, it is not expected to rollback, unless some crash happens). To do so, we write the index of the split index in a temporary file that we use until transaction finalization. We also keep a backup of the initial inline file to be able to rollback the split if needed. As a result, transaction rollback cancel the split and no longer corrupt fncache. We also no longer have a small inconsistency windows where the transaction could be unrecoverable.

fncache: make it possible to ignore some file In the next changeset, we need to able to ignore some temporary file. This changeset teach the fncache about that.

revlog: test that pending hooks properly see the repository on split This seems important to explicitly cover this case before changing the code.

revlog: test possible read race condition with splitting This is currently working fine, but could break with another approach (for example, with the one we are about to use in the next changesets…) So we make sure the case is covered.

revlog: add a failing variant of the the split + transaction test We have another variant to tests, and it is crashing… So lets cover it with tests.

revlog: update the split + transaction test We add section, increase the amount of comments and simplify some of the constructs. We are about to build more on top this tests so lets do a small cleanup first.

transaction: allow to backup file that already have an offset This will be useful in the next changeset to deal with rolling back the split of inline revlog on transaction failure. This involve deeper change to the transaction logic as we need to make sure we restore the backup -before- the truncation step. Otherwise, the truncation would act on the wrong file and be overwritten by the backup restoration later.

transaction: move the restoration of backup file in a small closure We are about to use that logic in two different location, making is a small reusable closure prepares that.

transaction: raise on backup restoration error A few line above, similar errors in the truncation code result in raising the associated exception. We should do the same here. This means the transaction recover is more strict now, which might be a problem when running `hg recover` in a share different from the one where the transaction fails. However this has always been a problem and need to be be addressed independently.

transaction: add clarifying comment about why ignoring some error is fine It is less scary when explained.

transaction: properly clean up backup file outside of .hg/store/ Oops.

branching: merge stable into default

statprof: with Python 3.12, lineno is (more) often None test-profile.t failed with errors like: TypeError: %d format: a real number is required, not NoneType statprof.py already handled None values as -1 in some cases. Do the same in more cases.

py3: fix for Python 3.12 emitting SyntaxWarning on invalid escape sequences Mercurial became very noisy after https://github.com/python/cpython/commit/a60ddd31be7ff96a8189e7483bf1eb2071d2bddf , for example: $ python3.12 mercurial/store.py mercurial/store.py:406: SyntaxWarning: invalid escape sequence '\.' EXCLUDED = re.compile(b'.*undo\.[^/]+\.(nd?|i)$') This verbosity made some tests fail. The problems were mostly insufficiently escaped regexps, relying on the Python parser/scanner preserving invalid escape sequences.

cext: fix for PyLong refactoring in CPython 3.12 Compiling Mercurial with Python 3.12 a5 would fail with: mercurial/cext/dirs.c: In function '_addpath': mercurial/cext/dirs.c:19:44: error: 'PyLongObject' {aka 'struct _longobject'} has no member named 'ob_digit' 19 | #define PYLONG_VALUE(o) ((PyLongObject *)o)->ob_digit[0] | ^~ mercurial/cext/dirs.c:97:25: note: in expansion of macro 'PYLONG_VALUE' 97 | PYLONG_VALUE(val) += 1; | ^~~~~~~~~~~~ mercurial/cext/dirs.c:19:44: error: 'PyLongObject' {aka 'struct _longobject'} has no member named 'ob_digit' 19 | #define PYLONG_VALUE(o) ((PyLongObject *)o)->ob_digit[0] | ^~ mercurial/cext/dirs.c:108:17: note: in expansion of macro 'PYLONG_VALUE' 108 | PYLONG_VALUE(val) = 1; | ^~~~~~~~~~~~ mercurial/cext/dirs.c: In function '_delpath': mercurial/cext/dirs.c:19:44: error: 'PyLongObject' {aka 'struct _longobject'} has no member named 'ob_digit' 19 | #define PYLONG_VALUE(o) ((PyLongObject *)o)->ob_digit[0] | ^~ mercurial/cext/dirs.c:145:23: note: in expansion of macro 'PYLONG_VALUE' 145 | if (--PYLONG_VALUE(val) <= 0) { | ^~~~~~~~~~~~ This was caused by https://github.com/python/cpython/commit/c1b1f51cd1632f0b77dacd43092fb44ed5e053a9 .

histedit: fix diff colors The problem here is that indexing a bytestring gives you integers, not chars, so the comparison to b'+' ends up being wrong. We don't really have a way to test curses output, so no tests to verify the correctness of this behaviour.

dirstate: fix a potential traceback when in `copy` and `rename` Before this changes, calling `hg copy` or `hg rename` could trigger a traceback about using an invalidated dirstate. This wasn't caught by the test as it needed the blackbox extension to preload the dirstate first in a way "refresh" invalidates it. Changing the context creation patterns fixes it.

dirstate: fix the bug in [status] dealing with committed&ignored directories In particular, these directories can "infect" their sibling directories with ignored status due to using a shared memoization cell by accident. This fixes bug #6795.

tests: demonstrate a bug with committed&ignored dirs

rust: remove out-of-date comment We've migrated to a newer version of Rust, so it doesn't make sense anymore.

rust: upgrade `rayon` dependency This includes a potential soundness fix as well as some improvements to performance which should be helpful.

rust: update zstd dependency Let's try to be the most up-to-date for this cycle. Fedora already has this version packaged, it's an added bonus.

fastexport: rework newline logic Newlines should only be added when otherwise the stream would look weird without them (on blobs), therefore they are the exception. Flip the logic so they are added, not skipped.

tests: simplify a bit

dirstate-v2: fix an incorrect handling of readdir errors Make sure not to cache the results of a failed readdir call.

tests: demonstrate a bug in dirstate-v2 handling of errors

tests: add a rewriting step to detect EACCES errors

encoding: avoid quadratic time complexity when json-encoding non-UTF8 strings Apparently the code uses "+=" with a bytes object, which is linear-time, so the whole encoding is quadratic-time. This patch makes us use a bytearray object, instead, which has a(n amortized-)constant-time append operation. The encoding is still not particularly fast, but at least a 10MB file takes tens of seconds, not many hours to encode.

bundle: prevent implicite bundling of internal changeset Now that the two mains source of on-disk bundle are preventing the inclusion of internal changesets in their bundling. We can add a lower level check that would prevent any other leakage of internal-phase changesets. We keep the door open to some usage, like the temporary bundle using during strip for example.

bundle: abort if the user request bundling of internal changesets See inline comments for details.

strip: do now include internal changeset in the strip backup See inline comment for details.

revset: add `_internal()` predicate This predicate help core code to select internal changeset (in the internal phase).

strip: drop the assert about bundled revision This is working well, so no need for extract computation. See the previous changeset for the rationnal of still have these assert around for a short interval.

strip: explicitly compute the boundary of the backup bundle We want to make change to the set of backed up revision in a future changeset, we start with a change of the computation without any changes in the semantic to clarify later changeset. The could of costly assert are here to testify that the result is still correct. They will be removed in the next changesets, but I wanted them in this changeset to help in case someone bisect a regression to this changeset in the future.

outgoing: fix common-heads computation from `missingroots` argument When initializing a `outgoing` object, the `common set` can be defined explicitly (with the `commonheads` argument`) or implicitly (with the missingroots arguments). It turns out the logic to compute `commonheads` from `missingroots` is buggy, as it does not consider the parents of enough changesets. Previously, it only considered parents of "missingroots` items, while it need to consider all parents of missing. Here is an example: F |\ C E | | B D |/ A If we use [E] as missing-roots, the missing set is [E, F], and the common-heads are [C, D]. However you cannot only consider the parent of [E] to find them, as [C] is not a parent of [E]. This already fix the bundle generated in one test, and it would prevent many other to misbehave with future change from this series.

bundle: include required phases when saving a bundle (issue6794) We now properly computes and includes phases above secret in bundle, previously, they would be skipped, and then the code computing them would crash. Note that from this changeset, we also include the heads associated with the changegroup's "target" phase. This turned out to be necessary to ensure the movement of changeset included in the bundle, but already known locally. This explain why lines for "secret" heads appears in multiple tests.

bundle: add test bundling changeset in the "archived" phase Having this test highlight that phase bundling above secret is significantly broken. Fixes coming in the next changesets.

bundle: introduce a "v3" spec This is a small update from "v2", we use the freshly out of experimental changelog version "03' format (alignment with bundle version is a coincidence) and we now bundle phases by default. We shall use v3 as the default bundle type later, when changegroup version '03' is a bit more established

bundle: add some phase boundary in the bundle type test case Same logic as the previous one, we want the tests to cover richer cases. It actually reveal a bug in `hg bundle foo.hg REMOTE` involving secret. So this is definitly not a bad idea.

bundle: expand the graph we us for bundletype/bundlespec tests We are about to test more things, especially phase bundling, so we need a graph a bit more complex than a single node. The test "code" was a bit simplified in the process, but no test-semantic changes were made.

revset: include all non-public phases in _notpublic We forgot up to update this when new phases were added.

dirstate: remove _actual_file_name methode This is dead-code

undo-files: cleanup legacy files when applicable The "journal" code is much more compact in 6.4, and so is the "undo" files as a result. However the previous version were much noisier, so let us cleanup undo files from older version too.

undo-files: clean existing files up before writing new one the in the initial design of journal/undo interaction, ages ago, new file always overwrote previous files. This is no longer the case for a long while, so it is time to properly clean things up before writing new ones. Otherwise, inconsistent "undo" state might exist on disk, leading `hg rollback` to misbehave (more that intended).

undo-files: make the undo-prefix configurable in `cleanup_undo_files` The transaction is configuration undo prefix, so we "need" it too.

undo-files: no longer pass the `repo` to `cleanup_undo_files` As foretold in the previous changesets, we no longer need a full repository object here.

undo-files: relies on a explicit list of possible undo files Instead of infering the list of undo files from the `_journalfiles` method on `localrepository`, we explicitly have the list of file in a constant next to the cleanup code. In practice this does not change much as `_journalfiles` is already returning the same "static" list and no internal or extensions extensions seems to actually wrap that. In addition, that list is not "too short" for cleanup, in case we need to cleanup undo files from older version of Mercurial that used to use more of them. this will be dealt with in a later changesets. This change is a step toward our goal to use the `cleanup_undo_files` within the transaction. The transaction has no reference to the `repo` object, so we need to move toward `cleanup_undo_files` not having one either.

undo-files: move the undo cleanup code in the transaction module Now that undo creation is gathered in the transaction module, let us move the code cleaning them up there too. This will be useful to better clean previous undo files up before creating new ones.

undo-files: drop the old undo rename logic It is no longer necessary I am not changing the transaction.__init__ signature since we are on stable right now.

undo-files: have the transaction directly tracks and manages journal rename This is much simpler this way.

undo-files: add a undoname closure to the _write_undo method We will also needs it when the transaction will take care of the other journal files, which is soon™.

undo-files: cleanup backup when cleaning undos Previously, the backups were left behind, by operation cleaning the undo's like strip, narrow and stream clone. The remaining elevant in the room is the transaction itself, who does not properly cleanup undo backup before copying the new ones.

undo-files: factor the vfs map in a repository property We define it in multiple locations and inconsistencies are appearing. So we now have a single definition point.

undo-files: add a utility function to read the backup-files definition We will need it in multiple places. so lets factor the logic around.

undo-files: use the cleanup function in streamclone Lets use the same code, so that we can fix things only once.

undo-files: also remove the undo.backupfiles The undo.backupfiles is dealt is directly managed by the transaction instead of going through the `localrepo.undofiles`. We start doing minimal management for it before using `cleanup_undo_files` on more situation. Proper handling of it is an intermediate goal of this series.

undo-files: use the cleanup function in narrow Lets use the same code, so that we can fix things only once.

undo-files: extract the cleanup code from strip in a function This logic is duplicated in multiple places and it missing some important parts. So lets start dealing with the duplication first.

run-tests: fix a crash when using the coverage options 35bf7f23b84c attempted to transition away from `distutils`, but the `packaging` code lacks `StrictVersion`. I have no idea when `packaging.version` became available, but I have it in python 3.6, so that should be good enough. For some reason, the import checker thinks this is a local import, and needs help to decide otherwise. Alternately we could ditch the version check entirely, because `coverage` is currently at 7.2.1, and the original check was added back in 2010.

rust: fix building on macOS (issue6801) The VFS change is copied over from Cargo, and likely to apply to other platforms as well. The dirstate change is essentially a replay of 440972d2175d, which was reverted in e98fd81bb151, part of !383, to silence some clippy warnings.

tests: fix timeout adjustement in delaypush.py Doing integer arithmetic with string is bound to fail.

revlog: add an exception hint when processing LFS flags without the extension It would be even better if this was either detected sooner, or the transaction completed (especially since the read/write processors aren't needed for the exchange). But this makes it easier for the user to resolve until that can be figured out.

tests: drop py36 conditionals in test-bad-extension.t Since this is a `>=` test, it's really conditionalizing py27 content, which isn't a thing anymore.

tests: drop py36 conditionals in test-hook.t Since this is a `>=` test, it's really conditionalizing py27 content, which isn't a thing anymore.

tests: drop py36 conditionals in test-http-bad-server.t Since this is a `>=` test, it's really conditionalizing py27 content, which isn't a thing anymore.

configitems: enable changegroup3 by default (unless using infinitepush) The LFS extension requires this, and if it isn't enabled on the client (or the LFS extension isn't loaded), a web client gets a 500 instead of a sensible error message. Now it gets a different (client) error, but maybe it can be handled more gracefully. c0f11347b107 indicates that treemanifest repos have this issue too. 29cfc474c5fd mentions gating this behind `experimental` so that the format could change, but that was 7 years ago and we now have an experimental `changegroup4` as well. We can keep this as a config for the next cycle in case someone runs into an unexpected problem, and then jettison it if the infinitepush bundle name changes are either acceptable as-is or can be created differently.

infinitepush: opt out of changegroup3 unless explicitly configured This is currently a no-op, as changegroup3 is disabled by default. But when it is enabled, it changes the hash names of the bundle files. As I don't use this extension, I have no idea if that's OK or not. So keep the current default behavior until we can get more info from actual users, while allowing them to opt-in for testing purposes.

statichttprepo: unbyteify several IOError messages Builtin errors generally want str messages.

statichttprepo: fix the vfs.join() method to match the base class definition Flagged by PyCharm. The superclass implementation raised NotImplementedError.

mail: unbyteify the SMTPException message This is a subclass of OSError, which wants str instead of bytes.

mail: add a missing argument to properly override starttls I didn't look into when this changed, but it was flagged by PyCharm.

crecord: switch a curses argument to bool to appease type checkers

grep: avoid reassigning byteskwargs to strkwargs PyCharm flagged each of these `get()` calls with bytes. We still pass the bytes form to the formatter to avoid changing the API, until all callers can be changed.

cat: drop unnecessary internal roundtrip of kwargs PyCharm seems to stick with the type at the initial assignment, so it flagged the `get()` with a bytes key since the method argument has str keys. It wasn't a bug, but then the bytes form is otherwise unused, and converted back to str.

relnotes: add 6.4 and empty next

Added signature for changeset 05de4896508e

Added tag 6.4rc0 for changeset 05de4896508e

branching: merge default into stable

transaction: drive the aberratant branch special case away shoo shoo shoo shoo. Happy to remove this awful special case (that I introduced myself last week…)

transaction: remove the `branch` backup for transaction We can now back it up at the end of the transaction as we do for the rest of the dirstate.

dirstate: deprecate calling `setbranch` without a transaction parameter The new way is now enforced.

branch: pass current transaction when writing branch for transaction backup This will requires more change soon (as we can simplify this backup). This will be done in later changesets.

branch: pass current transaction when writing branch in shelve

branch: pass current transaction when writing branch in import

branch: pass current transaction when writing branch in backout command

branch: pass current transaction when writing branch in branch command

branch: pass current transaction when writing branch in merge

branch: pass current transaction when writing branch in rebase