packaging: upgrade packages in Windows environment I ran `pip-compile -U` to update all packages to latest versions. I did this so the Windows environment is more modern. dulwich 0.20 dropped support for Python 2.7. Rather than add yet another requirements.txt file, I decided to just pin at an older release to minimize complexity. Differential Revision: https://phab.mercurial-scm.org/D9187

contrib: install Python 3.9.0 We update the Windows and Linux environments to install CPython 3.9.0, which was released a few days ago. Differential Revision: https://phab.mercurial-scm.org/D9186

contrib: stop installing Python 3.5 and 3.6 in Windows environment We're only publishing Python 3.7+ wheels and other builds on Windows. Python 3.5 and 3.6 are not supported on this platform. Let's stop installing them in the development environment. Differential Revision: https://phab.mercurial-scm.org/D9185

py3: convert an exception message to bytes One of our users ran into this (unrelated to my recent work on more specific errors). Differential Revision: https://phab.mercurial-scm.org/D9184

tests: fix test-url.py on py3, broken by D9179 Differential Revision: https://phab.mercurial-scm.org/D9180

revset: add a `node` key for sorting Sorting by node give an arbitrary, but stable order which is useful for sampling.

pyoxidizer: produce working Python 3 Windows installers (issue6366) While we've had code to produce Python 3 Windows installers with PyOxidizer, we haven't been advertising them on the web site due to a bug in making TLS connections and issues around resource handling. This commit upgrades our PyOxidizer install and configuration to use a recent Git commit of PyOxidizer. This new version of PyOxidizer contains a *ton* of changes, improvements, and bug fixes. Notably, Windows shared distributions now mostly "just work" and the TLS bug and random problems with Python extension modules in the standard library go away. And Python has been upgraded from 3.7 to 3.8.6. The price we pay for this upgrade is a ton of backwards incompatible changes to Starlark. I applied this commit (the overall series actually) on stable to produce Windows installers for Mercurial 5.5.2, which I published shortly before submitting this commit for review. In order to get the stable branch working, I decided to take a less aggressive approach to Python resource management. Previously, we were attempting to load all Python modules from memory and were performing some hacks to copy Mercurial's non-module resources into additional directories in Starlark. This commit implements a resource callback function in Starlark (a new feature since PyOxidizer 0.7) to dynamically assign standard library resources to in-memory loading and all other resources to filesystem loading. This means that Mercurial's files and all the other packages we ship in the Windows installers (e.g. certifi and pygments) are loaded from the filesystem instead of from memory. This avoids issues due to lack of __file__ and enables us to ship a working Python 3 installer on Windows. The end state of the install layout after this patch is not ideal for @: we still copy resource files like templates and help text to directories next to the hg.exe executable. There is code in @ to use importlib.resources to load these files and we could likely remove these copies once this lands on @. But for now, the install layout mimics what we've shipped for seemingly forever and is backwards compatible. It allows us to achieve the milestone of working Python 3 Windows installers and gets us a giant step closer to deleting Python 2. Differential Revision: https://phab.mercurial-scm.org/D9148

tests: set git config using `git config` for simplicity I also took the liberty to not set the same config value twice as we did before. Differential Revision: https://phab.mercurial-scm.org/D9176

tests: use `git init` instead of unusual synonym `git init-db` I had never heard of `git init-db` until I saw our tests. Differential Revision: https://phab.mercurial-scm.org/D9175

errors: name arguments to Abort constructor Differential Revision: https://phab.mercurial-scm.org/D9179

errors: stop passing non-strings to Abort's constructor The next patch will change `Abort`'s constructor and `__bytes__` functions and they will start assuming that the first argument is the messages as `bytes`. Differential Revision: https://phab.mercurial-scm.org/D9178

errors: name arguments to UnknownCommand constructor Differential Revision: https://phab.mercurial-scm.org/D9166

errors: name arguments to AmbiguousCommand constructor Differential Revision: https://phab.mercurial-scm.org/D9165

errors: name arguments to CommandError constructor Because readability counts. Differential Revision: https://phab.mercurial-scm.org/D9164

errors: move UnknownCommand and AmbiguousCommand near CommandError They seem closely related. Differential Revision: https://phab.mercurial-scm.org/D9163

templatekw: make {successorssets} always return a list (issue6342) Previously, {successorssets} returns an empty string instead of an empty list for a non-obsolete changeset. The changing type of the JSON output makes it hard to consume from statically-typed languages. Differential Revision: https://phab.mercurial-scm.org/D9158

tests: test output of {predecessors}, {successorssets} for non-obsolete csets Differential Revision: https://phab.mercurial-scm.org/D9157

branching: merge with stable

copies: move `merged` testing sooner previously `is_merged` was an expensive callback. Now that all this data is pre-computed, this is is a simple membership testing. So it is probably cheaper than the membership testing. Differential Revision: https://phab.mercurial-scm.org/D9141

copies: return None instead of ChangingFiles when relevant If they are not relevant data, they are no need to fetch them. Differential Revision: https://phab.mercurial-scm.org/D9140

copies: add a HASCOPIESINFO flag to highlight rev with useful data If some files changes that may impact copy tracing are detected, we set this flag. This helps the copy tracing algorithm to skip fetching possibly expensive data when unnecessary. Differential Revision: https://phab.mercurial-scm.org/D9139

salvaged: properly deal with salvaged file during copy tracing When salvaged files are encountered, the removal have been reverted and we should keep the rename information from the other side. The conditional was starting to be quite hairy, so we spell it out in multiple elif case for readability. This fixes the associated test cases introduced a while back. The changeset centric copy tracing is now (known) bug free. Differential Revision: https://phab.mercurial-scm.org/D9130

salvaged: persist the salvaged set on disk With the new side data storage, this is trivial. Differential Revision: https://phab.mercurial-scm.org/D9129

changing-files: add clean computation of changed file for merges This is "a tad more complicated" than the previous cases. See inline documentation for details (have fun). Differential Revision: https://phab.mercurial-scm.org/D9128

changing-files: add clean computation of changed files for linear changesets The `files` field is not reliable, so we need to compute things from scratch. We deal with the second simplest case, linear changesets. We diff the current manifest with the parent manifest. This reveal the file added, changed and removed. Differential Revision: https://phab.mercurial-scm.org/D9127

changing-files: add clean computation of changed files for roots The `files` field is not reliable, so we need to compute things from scratch. We start with the simplest case root changesets. In the beginning they was nothing, then user said "let there be files" and there were added files. Differential Revision: https://phab.mercurial-scm.org/D9126

changing-files: add a debug command display changed files The binary output from sidedata is useful to verify the underlying data do not get corrupted. However having a human readable version is much simpler for debuging the changed files data itself. So we add a debug command to dump this information and we use it in the tests. Differential Revision: https://phab.mercurial-scm.org/D9125

side-data: add a test to check sidedata upgrade We want to make sure that sidedata computed at commit time and during upgrade are the same. To do so, we upgrade the repository in `test-copies-chain-merge.t` since the file is checking all the special case that we need to catch. The sidedata content in both case should be the same. Differential Revision: https://phab.mercurial-scm.org/D9124

changing-files: split the changing files computation from encoding The side data encoding is already in its own function, now we move the changing files computation in it own function, it will receive more updates in the coming changesets. Differential Revision: https://phab.mercurial-scm.org/D9123

salvaged: record salvaged in ChangingFiles at commit time The new code is a simple but effective way to detect this information. We might be able to move it inside the various conditionnal above, but I want to focus on simplicity until we have a full working stack. It is worth noting that if we record the information in the ChangingFiles object, it is not persisted yet. This will comes with later changesets. Differential Revision: https://phab.mercurial-scm.org/D9120

salvaged: track removal-candidates in more cases If we want to use this information for copy tracing, then we need to be able to record it. First we unlock the recording of deletion candidates, and we will actually record the data in the next changeset. Differential Revision: https://phab.mercurial-scm.org/D9119

salvaged: explicitly skip salvaged file while encoding This is a temporary measure to avoid diverging test cases (between commit time sidedata and upgraded sidedata) while we implement salvaged tracking in the different cases. This will be dropped soon. Differential Revision: https://phab.mercurial-scm.org/D9122

changing-files: add a "salvaged" set to track file that were not removed We need this set for the copy tracing algorithm. See documentation for details about this set. Differential Revision: https://phab.mercurial-scm.org/D9118

tests: back out accidentally queued changeset 0627cd03b1e9 I accidentally queued other people's patches on top of my own and pushed mine along with theirs.

tests: fix test-check-module-imports.t broken by D9150 Differential Revision: https://phab.mercurial-scm.org/D9177

tests: make test-convert-git.t work across changed default branch name Recent git versions default to "main" instead of "master", which breaks our test. Differential Revision: https://phab.mercurial-scm.org/D9174

tests: update test-copies-chain-merge.t to not use empty files Merging empty files is not very interesting or realistic.

tests: only diff the `Cargo.lock` in `test-check-cargo-lock.t` While not perfect (you could be running tests while developing with a justified modification to `Cargo.lock`), that use-case is small enough that I don't think it's really going to be an issue. This stops the test from failing for unrelated changes when running it locally. Differential Revision: https://phab.mercurial-scm.org/D9170

tests: update/remove outdated comments in test-copies-chain-merge.t 1b8fd4af3318 fixed the test case but missed updating the comments mentioning the bug. Thanks for @martinvonz who spotted these outdated comments. Differential Revision: https://phab.mercurial-scm.org/D9168

churn: leverage logcmdutil to filter revisions by --date

cmdutil: remove remainder of old walkchangerevs() implementation

cmdutil: rewrite walkchangerevs() by using logcmdutil functions cmdutil.walkchangerevs() now takes (revs, makefilematcher) in place of (match, opts), and only provides the "windowing" functionality. Unused classes and functions will be removed by the next patch. "hg grep --follow" (--all-files) is still broken since there is no logic to follow copies while traversing changelog, but at least, it does follow the DAG.

grep: filter target files by matcher Prepares for the migration to logcmdutil's logic, where cmdutil.walkchangerevs() won't always build an exact set of paths to be scanned.

cmdutil: make walkchangerevs() call prepare with matcher instead of filenames Prepares for migrating walkchangerevs() to logcmdutil's logic, which provides matcher-based interface.

grep: add option for logcmdutil.makewalker() to not filter revs by file pats This is needed to implement "grep --all-files".

rust-parsers: use in-place mutation instead of allocating a new `Vec` This is not done for the `dirstate-tree` feature, since it lacks `iter_mut`. Differential Revision: https://phab.mercurial-scm.org/D9136

revlog: pre-compute null revlog item for pure version The dynamically created tuple shows up in memory profiles and the use of nullid prevents the normal constant building to work. Differential Revision: https://phab.mercurial-scm.org/D9154

packaging: normalize - to _ in WiX Id values - is not a valid identifier character in WiX Ids. So let's normalize accordingly. I discovered this issue after a subsequent change which introduces a directory with a - in its name. Differential Revision: https://phab.mercurial-scm.org/D9147

contrib: install Rust 1.46.0 Let's use the latest/greatest version of Rust in the automation environment. Differential Revision: https://phab.mercurial-scm.org/D9146

contrib: install latest versions of Python Let's keep our build environment modern. We upgrade 3.5-3.8 on Linux. Just 3.7 and 3.8 on Windows because we don't use <3.7 on Windows. Differential Revision: https://phab.mercurial-scm.org/D9145

help: extract logic for listing commands and topics Differential Revision: https://phab.mercurial-scm.org/D9134

copies: directly pass a changes object to the copy tracing code The object contains all the data we need. For example, the `is_merged` callback can now use the associated precomputed data. This will be useful again soon when the `salvaged` set will be introduce to solve the issue with delete file reverted during a merge. See 4b582a93316a and 14be07d5603c for details. Differential Revision: https://phab.mercurial-scm.org/D9117

copies: no longer change the sidedata flag With the new sidedata storage that include data about all file changes, every revision has one, so the sidedata flag is not longer a good way to spot changeset with copy information. So we drop this check to simplify the code We optimisation itself provided an interesting speedup, so we will likely reintroduce something similar, with a dedicated flag, in the future. Differential Revision: https://phab.mercurial-scm.org/D9116

copies: use dedicated `_revinfo_getter` function and call We want to return data in a different form, so we need different revinfo function. At that point it make sense to have different getter. Differential Revision: https://phab.mercurial-scm.org/D9115

copies: make two version of the changeset centric algorithm They are two main ways to run the changeset-centric copy-tracing algorithm. One fed from data stored in side-data and still in development, and one based on data stored in extra (with a "compatibility" mode). The `extra` based is used in production at Google, but still experimental in code. It is mostly unsuitable for other users because it affects the hash. The side-data based storage and algorithm have been evolving to store more data, cover more cases (mostly around merge, that Google do not really care about) and use lower level storage for efficiency. All this changes make is increasingly hard to maintain de common code base, without impacting code complexity and performance. For example, the compatibility mode requires to keep things at different level than what we need for side-data. So, I am duplicating the involved functions. The newly added `_extra` variants will be kept as today, while I will do some deeper rework of the side data versions. Long terms, the side-data version should be more featureful and performant than the extra based version, so I expect the duplicated `_extra` functions to eventually get dropped. Differential Revision: https://phab.mercurial-scm.org/D9114

changing-files: retrieve changelogrevision.files from the sidedata block The `files` field is know to have issue, using a list with fixed, and fixable, computation can only help. For example, using a fixes `files` field would be enough to fix issue6219 once this feature get usable in production. We focus on having thing working for now, we will deal with performance later. Right now we have a ironic situation were we parse sorted value from disk to turn them into a set and then having to sort it again. Differential Revision: https://phab.mercurial-scm.org/D9092

changing-files: drop the now useless changelogrevision argument Since all filename are now included in the sidedata block, we no longer need to decode the `files` from the revision. Differential Revision: https://phab.mercurial-scm.org/D9091

changing-files: rework the way we store changed files in side-data We need to store new data so this is a good opportunity to rework this fully. 1) We directly store the list of affected file in the side data: * This avoid having to fetch and parse the `files` list in the revision in addition to the sidedata. Making the data more self sufficient. * This work around situation where that `files` field contains wrong information, and open the way to other bug fixing (eg: issue6219) * The format (fixed initial index, sorted files) allow for fast lookup of filename within the structure. * This unify the storage of affected files and copies sources and destination, limiting the number filename stored redundantly. * This prepare for the fact we should drop the `files` as soon as we do any change affecting the revision schema. * This rely on compression to avoid a significant increase of the changelog.d. More testing on this will be done before we freeze the final format. 2) We can store additional data: * The new "merged" field, * A future "salvaged" set recording files that might have been deleted but have were still present in the final result. Differential Revision: https://phab.mercurial-scm.org/D9090

tests: skip doctests if not running from a hg repo Differential Revision: https://phab.mercurial-scm.org/D9150

py3: use native string when comparing with a function's argspec I only found two such bugs in `contrib/perf.py` Differential Revision: https://phab.mercurial-scm.org/D9149

test: try to unflaky test-profile.t That test rely on timing measurement, because it is about timing measurement. We try to filter out the most common source of flakyness (slow disk stating)

merge: check for conflicting actions irrespective of length of bids We should for whether bids contain a combination of actions which conflict with each other. Since right now we only have couple of such combination, and combinations also consist of two actions, we were checking for them only when length of bids is 2. Let's check that irrespective of the length of bids.

branching: merge with stable

log: extract function that builds (revs, makefilematcher) from walkopts "hg grep" and "hg churn" will use this interface.

log: move --graph and topo sort options to walkopts This is the last opts.get() found in getrevs(). It might be better to define an enum, but for now, it is just a string.

log: move miscellaneous filter/matcher options to walkopts Maybe this is the most uninteresting patch. I want to remove the dependency on untyped/unlabeled opts dict at all. I thought --no-merges and --only-merges could be merged into a single enum value, but actually they both can be set to filter out everything.

log: map --removed to walkopts.force_changelog_traversal This is the flag to forcibly enable the slowpath. I'm not sure if the slowpath parameter should be merged with this flag, so let's keep it as an immutable flag for now. I'll add another flag to support "grep --all-files". These two will be the flags which aren't directly mapped from the command-line options.

copies: rename some function to the new naming scheme We are about to introduce more function and more variants, it seems better to align everything on the new scheme, allowing `_` for clarify. We do this aligned in a dedicated changeset to make the next changesets simpler. Differential Revision: https://phab.mercurial-scm.org/D9113

changing-files: cache the various property We are going to start using them more having some basic caching would be good. The focus is not about performance yet, however avoid some easy N² trap seems reasonable. Differential Revision: https://phab.mercurial-scm.org/D9112

changing-files: always use `mark_touched` to update the touched set We use this function internally too because that will make cache invalidation simpler. Differential Revision: https://phab.mercurial-scm.org/D9111

changing-files: fix docstring As pointed by Pulkit Goyal. Differential Revision: https://phab.mercurial-scm.org/D9143

rust: move rustfmt.toml to repo root so it can be used by `hg fix` `hg fix` runs the formatters from the repo root so it doesn't pick up the `rustfmt.toml` configs we had in each the `hg-core`, `hg-cpython`, and `rhg` packages, which resulted in warnings about `async fn` not existing in Rust 2015. This patch moves the `rustfmt.toml` file to the root so `hg fix` will use it. By putting the `rustfmt.toml` file in a higher-level directory, it also applies to the `chg` and `hgcli` packages. That makes `test-check-rust-format.t` fail, so this patch also applies the new formatting rules to those packages. Differential Revision: https://phab.mercurial-scm.org/D9142

merge: if CHANGED_DELETED and KEEP_NEW are actions, choose CHANGED_DELETED ACTION_KEEP_NEW and ACTION_CHANGED_DELETED are conflicting actions as one says that file is new while other says that file was present earlier and has changed-delete conflicts. Let's do changed-delete which will lead to conflicts and make user choose the right way.

tests: fix a typo in `desc()` revset in test-merge-criss-cross.t I was looking into fixing the broken behavior only to find out that we had a typo in the commit message and merge was not performed.

merge: if DELETED_CHANGED and GET are in actions, choose DELETED_CHANGED ACTION_GET represents that either the file is created on remote or it's newer on the remote side. However, since we have a ACTION_DELETE_CHANGED too, it means the file is not present locally and ACTION_GET is representing that file was created on remote. Having both ACTION_GET and ACTION_DELETED_CHANGED is conflicting because one says that file was created on remote and other says file has delete-changed conflicts. Let's choose ACTION_DELETED_CHANGED which will result in conflicts and make user choose the right way forward.

tests: add newfilenode test case in test-merge-changedelete.t

rust: leverage improved match ergonomics Differential Revision: https://phab.mercurial-scm.org/D9137

rust: update minimum supported Rust compiler version Debian stable backported `rustc 1.41.1` for a Firefox release that needed it, so we can finally update the compiler version to something that isn't more than a year old. This means that (among other things [1]), `async await` syntax is usable! Differential Revision: https://phab.mercurial-scm.org/D9138

changing-files: add a utility to compute the merged files post-commit We will need it in `_getsidedata` as soon as we start persisting that set. Differential Revision: https://phab.mercurial-scm.org/D9089

changing-files: record merged files at commit time The data is easy to gather at commit time, and we need it for changeset centric copy tracing. Right now, it is not persisted so we cannot use it. However we will fix this part very soon, gathering something to persist was necessary first. Differential Revision: https://phab.mercurial-scm.org/D9088

changing-files: add the ability to track merged files too The set of merged files is used when doing changeset centric copy tracing (cf `is_merged` in `mercurial/copies.py`. So tracking (and persisting) this set will be useful. We start with adding the attribute on the new object. Differential Revision: https://phab.mercurial-scm.org/D9087

rust: start plugging the dirstate tree behind a feature gate The previous patch added the `dirstate-tree` feature gate to enable the two dirstate implementations to co-habit while the tree-based one gets better. This patch copies over the code that differs, be it because the algorithm changed or because the borrowing rules are different. Indeed, `DirstateTree` is not observationally equivalent to the std `HashMap` in the APIs we use: it does not have the `Entry` API (yet?) and its iterator returns owned values instead of references. This last point is because the implementation needs to be changed to a more clever and efficient solution. Differential Revision: https://phab.mercurial-scm.org/D9133

rust: introduce `dirstate-tree` cargo feature This feature gates (at compile-time) the use of the newly-added dirstate tree. The motivation for this is that the dirstate tree is currently *very* slow; replacing the current hashmap-based dirstate is not a viable solution in terms of performance... and why would you be using the Rust implementation if not for performance? The feature will also help reviewers better understand the differences that will slowly appear as the dirstate tree gets better. Differential Revision: https://phab.mercurial-scm.org/D9132

rust: fix formatting Maybe there is a small divergence in the nightly formatter? This didn't seem to trip the CI. Differential Revision: https://phab.mercurial-scm.org/D9131

rust-matchers: make `Matcher` trait object-safe Before this patch, it is not possible to create a `Matcher` trait-object (like `Box<dyn Matcher>`), because of the use of a generic parameters in some methods, namely `impl AsRef<HgPath>`. While this makes the interface less flexible for callers in theory, it does not change anything in the current codebase. Until something like [1] is implemented, this is a "tradeoff" that we need to make anyway. [1] https://internals.rust-lang.org/t/pre-rfc-expand-object-safety/12693 Differential Revision: https://phab.mercurial-scm.org/D9071

rust: clippy pass This removes some justified warnings and one hard error that, while technically not a bug, was an ugly oversight on my part. Differential Revision: https://phab.mercurial-scm.org/D9094

updaterepo: add deprecation warning We just moved all in-tree callers off of the function, but we're leaving it here for a bit to help extension writers. Differential Revision: https://phab.mercurial-scm.org/D9110

rhg: use `.or(Err(Error))` not `.map_err(|_| Error)` (D9100#inline-15067) Differential Revision: https://phab.mercurial-scm.org/D9109

hg-core: use `.or(Err(Error))` not `.map_err(|_| Error)` (D9100#inline-15067) Differential Revision: https://phab.mercurial-scm.org/D9108

hg-core: return Err if `offset != bytes.len()` (D8958#inline-14994 followup 2/2) [X] make `Index` owner of its bytes [X] make `Index::new` return an error if `offset != bytes.len()` Differential Revision: https://phab.mercurial-scm.org/D9107

hg-core: make `Index` owner of its bytes (D8958#inline-14994 followup 1/2) Prevent building `Index` every time it is needed. It was a bad idea anyway. When `Index::new` will return `Result` it will avoid things like `Revlog::len` returning `Result<usize>` instead of `usize`. [X] make `Index` owner of its bytes [ ] make `Index::new` return an error if `offset != bytes.len()` Differential Revision: https://phab.mercurial-scm.org/D9106

hg-core: renaming of `Chunk` offset methods (D8958#inline-15002 followup) Differential Revision: https://phab.mercurial-scm.org/D9105

hg-core: minor rewording in docstring (D8958#inline-15005 followup) Differential Revision: https://phab.mercurial-scm.org/D9104

hg-core: use anonymous lifetime for `impl Chunk` (D8958#inline-15003 followup) Differential Revision: https://phab.mercurial-scm.org/D9103

hg-core: use `u32` instead of `i32` in `Chunk` (D8958#inline-15001 followup) Differential Revision: https://phab.mercurial-scm.org/D9102

hg-core: use the term `chunk` instead of `frag` (D8958#inline-15000 followup) Differential Revision: https://phab.mercurial-scm.org/D9101