clone: make sure we warm the cache after a clone This work around any deviciency/limitation of the clone process. In our case this ensure the persistent nodemap exist with valid content. Ideally, the cloning process would also do "the right thing". However since older server will never be able to do "the right thing". The local workaround will be necessary anyway. I am not worried by the performance impact of this as `hg clone` is non-instant on large repositories where is could matters. Warming the cache if they are already correct is very fast. And if they are not already warm, this seems like a good time to do so. This impact various test as more cache are now warmed sooner, all the change should be harmless. Differential Revision: https://phab.mercurial-scm.org/D9789

persistent-nodemap: remove the "experimental" tag to the feature This has been around for a while and has all the discussed configuration options. Differential Revision: https://phab.mercurial-scm.org/D9764

persistent-nodemap: drop the storage.revlog.nodemap.mode config We now have a new, better, higher level, option. So we can drop the older one as planned. Differential Revision: https://phab.mercurial-scm.org/D9763

persistent-nodemap: add a "abort" option to the slow-path config We make it the default, and document the behavior in the help for the main config option. Differential Revision: https://phab.mercurial-scm.org/D9762

persistent-nodemap: add a "warn" option to the slow-path config And make it the default until we get an abort option. Differential Revision: https://phab.mercurial-scm.org/D9761

persistent-node: check the value of the slow-path config We should probably provide some standard for this in config item, but this is a quest for another adventure. Differential Revision: https://phab.mercurial-scm.org/D9760

persistent-nodemap: add a revlog.storage.persistent-nodemap.slow-path option As discussed during the sprint, we want to prevent user to get an unexpected performance regression when accessing a repository using "persistent-nodemap" without the associated Rust extension. We start by adding a config declaration and some documentation. Since "allow" is the current behavior, we don't need to add any code. The option possible value will come later. Note that we already have a `storage.revlog.nodemap.mode` option, but that option is a bit different. It does some warning and checking at revlog instantiation time. While we want something done at requirements checking time. Since we plan for new names and new config value names, we introduce a new option and will drop the old one later. Differential Revision: https://phab.mercurial-scm.org/D9758

persistent-nodemap: document storage.revlog.persistent-nodemap.mmap=no The feature is gaining public facing document, its options too. Differential Revision: https://phab.mercurial-scm.org/D9757

persistent-nodemap: rename the storage.revlog.nodemap.mmap option Since the feature is publicly known as "persistent-nodemap" this seems better to reference it as "persistent-nodemap" in related config too. So we rename the option to `storage.revlog.persistent-nodemap.mmap`. Differential Revision: https://phab.mercurial-scm.org/D9756

persistent-nodemap: document the feature in `hg help config.format` This is necessary before we advertise the feature more. Differential Revision: https://phab.mercurial-scm.org/D9755

persistent-nodemap: write down the sprint conclusion This write down the plan for turning feature that are only fast when the rust extension exists. Future changesets will implement it. Differential Revision: https://phab.mercurial-scm.org/D9754

persistent-nodemap: document the `devel.persistent-nodemap` option What it does is not obvious. Differential Revision: https://phab.mercurial-scm.org/D9753

copies: fix some comment in _filter The scenario the comment describes match case 6, not case 2. Differential Revision: https://phab.mercurial-scm.org/D9793

discovery: move some debug output closer to were it belong I assume these debug output, increment and comment drifted over time. Differential Revision: https://phab.mercurial-scm.org/D9800

discovery: add a discovery.grow-sample.rate This allow to control the effect of the growth rate on the discovery process while doing analysis. Differential Revision: https://phab.mercurial-scm.org/D9799

discovery: add a `devel', b'discovery.grow-sample` That option make it possible to disable the "sample growing" behavior when doing analysis and comparison. Differential Revision: https://phab.mercurial-scm.org/D9798

discovery: document the `devel.discovery.randomize` option Gratuitous improvement as I was passing by this config section to add more. Differential Revision: https://phab.mercurial-scm.org/D9797

shelve: move listshelves() to new ShelfDir class, so caller need not pass vfs Differential Revision: https://phab.mercurial-scm.org/D9744

shelve: also create class representing whole directory of shelves It's a little annoying to have to create and pass in a vfs into `listshelves()`. This patch attempts to start addressing that by creating a class that represents a directory containing shelves (the directory can be either `.hg/shelved/` or `.hg/shelve-backup/`). Differential Revision: https://phab.mercurial-scm.org/D9743

shelve: add a method for deleting shelf to new shelf class This is not necessary for my future changes, but it's more consistent to encapsulate the knowledge of the various files in the `Shelf` class. Differential Revision: https://phab.mercurial-scm.org/D9742

shelve: inline ".patch" constant now that it's only used in the Shelf class Differential Revision: https://phab.mercurial-scm.org/D9741

shelve: use listshelves() in cleanupoldbackups() With this patch, there are no more assumptions outside the `Shelf` class about which files (`.patch`, `.hg`, `.shelve`) make up a shelf. As such, this finishes the preparations for making phase-based shelve (perhaps optionally) not write the `.patch` and `.hg` files. Differential Revision: https://phab.mercurial-scm.org/D9740

shelve: make listshelves() list shelves in a given vfs Differential Revision: https://phab.mercurial-scm.org/D9739

shelve: replace repo instance in Shelf class by vfs instance I'd like to be able to teach the `Shelf` class to read shelves from `.hg/shelve-backup/` in addition to `.hg/shelved/`. That means that we'll have to pass in a vfs instead of the repo, so this patch does that preparation. Differential Revision: https://phab.mercurial-scm.org/D9738

shelve: use listdir() instead of readdir() when we don't need stat information Differential Revision: https://phab.mercurial-scm.org/D9737

shelve: don't crash on file with unexpected extension in .hg/shelved/ We assumed that the files in the `.hg/shelved/` directory have an extension. That's a valid assumption except that users may put garbage in the directory. This patch fixes the crash by simply not assuming that the result of splittin a string at '.' yields an extension. We don't use the extension since the previous patch anyway. Differential Revision: https://phab.mercurial-scm.org/D9720

shelve: don't include invalid shelves in `hg shelve --list` Before this patch, if a shelved change is missing its `.hg` file, we still list it in `hg shelve --list`, but then `hg unshelve` crashes. This patch makes it so we only list valid shelved changes. This patch means that users who do `touch .hg/shelve/buy-milk.patch` as a form of TODO list will no longer see their TODO items in `hg shelve --list`. Differential Revision: https://phab.mercurial-scm.org/D9719

shelve: extract some repeated creation of shelf instances to variables This just looks cleaner to me; I'd be surprised if there's any measurable performance improvement. Differential Revision: https://phab.mercurial-scm.org/D9714

shelve: teach new shelf class to check if .shelve file exists This removes the only remaining use for `shelvedfile`, so the class now goes away. Differential Revision: https://phab.mercurial-scm.org/D9713

shelve: move method for creating backup to new shelf class Differential Revision: https://phab.mercurial-scm.org/D9712

shelve: make gennames() helper generate relative backup paths When I saw `gennames()`, I thought it was meant to take a relative filename as argument. Maybe it was or maybe it wasn't, but it seems simpler to pass it a relative path anyway, so that's what this patch does. That also makes the call to backupvfs.exists() more natural (I expect a relative path to be passed there too). Differential Revision: https://phab.mercurial-scm.org/D9711

shelve: use listshelves() in cleanup function The point of this patch is to make it so all the callers of `shelvedfile.movetobackup()` look the same, so I can move it over to the new `Shelf` class next. Differential Revision: https://phab.mercurial-scm.org/D9710

shelve: inline shelvedfile.filename() since there are no callers outside class Differential Revision: https://phab.mercurial-scm.org/D9709

shelve: make listshelves() return shelf names instead of filenames All three callers now prefer the shelf name over the filename (already before my recent patches, two out of three callers preferred the shelf name). Differential Revision: https://phab.mercurial-scm.org/D9708

shelve: move method for getting stat (mtime) to new shelf class Only the mtime was needed, so I made it restricted to that in the move. The new `Shelf` class expects its argument to be a shelf name (not a arbitrary filename like `shelvedfile` would accept), so only the shelf name is now passed in. Differential Revision: https://phab.mercurial-scm.org/D9707

shelve: open patch using new shelf class instead of open() For some reason the existing code didn't use `shelvedfile().opener()` so here we migrate to `shelf().open_patch()` from the `open()` system call instead. Differential Revision: https://phab.mercurial-scm.org/D9706

shelve: move function for opening .patch file to new shelf class The `opener()` method was used specifically for the `.patch` file, and the new `Shelf` class deals with all files involved in a shelf, so I renamed the function in the process. Differential Revision: https://phab.mercurial-scm.org/D9705

shelve: move method for reading .hg to new shelf class Differential Revision: https://phab.mercurial-scm.org/D9704

shelve: move method for writing bundle to new shelf class Differential Revision: https://phab.mercurial-scm.org/D9703

shelve: move method for reading .shelve file to new shelf class Differential Revision: https://phab.mercurial-scm.org/D9702

shelve: move method for writing .shelve to new shelf class Differential Revision: https://phab.mercurial-scm.org/D9701

shelve: introduce class representing a shelf I'm about to make phase-based shelve not write `.hg` and `.patch` files. Having a class that represents a single shelf, regardless of which files it uses will help. I'm starting small with just a `.exists()` function. I plan to eventually remove the `shelvedfile` class once all functionality has been moved to the new class. By the way, I know that things you shelve are not typically themselves shelves. I still picked `Shelf` for the class because it's short (compared to e.g. `ShelvedChange`). Differential Revision: https://phab.mercurial-scm.org/D9700

shelve: raise more specific errors Differential Revision: https://phab.mercurial-scm.org/D9699

shelve: trust caller of shelvedfile.opener() to check that the file exists The only place we call `shelvedfile.opener()` is when we're about to apply a bundle. The file should always exist. If it doesn't, the `.hg/` directory is corrupt and we don't provide any guarantees about supporting corrupt repos (besides, telling the user that the shelve doesn't exist when `hg shelve --list` lists it is not very helpful). Differential Revision: https://phab.mercurial-scm.org/D9698

shelve: rewrite check for unknown shelf to delete The code would try to delete the shelf's .patch file and if that raised an exception, it would convert it to an `error.Abort`. This patch rewrites it so the check is done upfront. I find it easier to read that way. It's now clear enough that I removed the comment explaining it as well. As Joerg pointed out during review, another differences is that the old code would move a `.hg` file without its `.patch` friend to backup before it realized that the `.patch` file was missing. The new code will error out earlier and not move the `.hg` file, which seems like an improvement. That should only matter on corrupt `.hg/shelved/` directories, however. Differential Revision: https://phab.mercurial-scm.org/D9697

shelve: remove a bundlerepo method It was added in 43816070284e (shelve: add a bundlerepo method, 2014-10-10), but I haven't been able to find a place it was used. Differential Revision: https://phab.mercurial-scm.org/D9696

tests: add tests for corrupt .hg/shelved/ directory I don't care much how we behave in these cases, except that we should provide a way for the user to get out of the broken state. Differential Revision: https://phab.mercurial-scm.org/D9718

ui: add a "config_default" method This allow code to access the expected value for a config. This is useful in the context of dynamic default value, and short term it will be useful to write dynamically "correct" code. Differential Revision: https://phab.mercurial-scm.org/D9759

test: split tests/test-merge-combination.t into multiple parts That test is quite slow, lets split it in smaller bits. Differential Revision: https://phab.mercurial-scm.org/D9768

test: extract the `genmerges` out of test-merge-combination.t This open the way to splitting this slow test in multiple ones. Differential Revision: https://phab.mercurial-scm.org/D9767

hghave: clarify `sqlite` requirements We need more than the python module, we also need the sqlite3 command line. Differential Revision: https://phab.mercurial-scm.org/D9787

localrepo: fix comment typo Differential Revision: https://phab.mercurial-scm.org/D9782

error: use detailed exit code 10 for command errors Command errors (unknown or ambiguous commands, or bad command arguments or options) are handled in the `dispatch` module. Most other errors are handled in the `scmutil` module. This patch therefore has to duplicate a little bit of code from the `scmutil` module. It's just a few lines, however, so it seems fine to me. It's a pretty common category of errors, so it's important to have them respect `ui.detailed-exit-code`. Differential Revision: https://phab.mercurial-scm.org/D9777

errors: raise InputError when non-existent help section requested Differential Revision: https://phab.mercurial-scm.org/D9776

revlog: migrate from PyEval_CallObject to PyObject_Call The former was deprecated in 3.9.0. Differential Revision: https://phab.mercurial-scm.org/D9791

fuzz: fix Makefile default PYTHON_CONFIG_FLAGS to be modern This is actually what we already do on oss-fuzz, so it's more correct as tests go. Differential Revision: https://phab.mercurial-scm.org/D9790

tests: wider work around matching in test-narrow-shallow.t Since issue6150 is still not fixed, we have to accommodate a larger array of possible results. Otherwise we get frequent flakiness of local and CI runs. Differential Revision: https://phab.mercurial-scm.org/D9766

fuzz: try and ensure fuzzer tests run against the right python-config Also only under python 3. Differential Revision: https://phab.mercurial-scm.org/D9752

contrib: remove testing for `dirstate-tree` Rust feature This feature will be replaced in a few months, and served more as a proof-of-concept. Keeping it in CI when no one is using it anymore is just wasteful. Differential Revision: https://phab.mercurial-scm.org/D9716

perf: don't turn byte to string when formatting perfbranchmap I am not sure why this `str` all is there is the first place. Differential Revision: https://phab.mercurial-scm.org/D9751

branchmap: avoid ancestor computations in absence of non-continous branches The branchhead computation is one of the more heavy operations for bigger repositories as it has to scan all changesets and potentially involves the expensive computation of the ancestor sets. Redo the computation to handle the common cases directly and use tighter conditions for when the ancestor scan is necessary. Most importantly, avoid it completely if the non-continous branches are processed in one update as seen in the initial computation after a clone. For the Mercurial repository, it gives a small 2-3% performance boost. For the NetBSD test repository, it cuts the time in half. Differential Revision: https://phab.mercurial-scm.org/D9631

persistent-nodemap: also list related file as part of the store This make sure they are will be selected during upgrade, and copy based clone. Differential Revision: https://phab.mercurial-scm.org/D9749

upgrade: explicitly test for revlog index We do not want to exclude `.d` we want to only include `.i`. We are about to add more extension type (for the persistent nodemap: `.n`, `.nd`) so lets make the list explicit instead. Differential Revision: https://phab.mercurial-scm.org/D9748

persistent-nodemap: highlight that node is not sent when streaming We have a local work-around, however having that data sent in the first place would be better. So we start with a test that monitor what is sent, and notice they are not. Differential Revision: https://phab.mercurial-scm.org/D9736

persistent-nodemap: test it (does not) exist after a stream clone This is currently buggy. We add a test before fixing the bug. Differential Revision: https://phab.mercurial-scm.org/D9734

persistent-nodemap: test it (does not) exist after a local clone This is currently buggy. We add a test before fixing the bug. Differential Revision: https://phab.mercurial-scm.org/D9733

persistent-nodemap: test that is it present after a clone The persistent nodemap is actually missing after local and stream clone. We start by adding a test for the case that works so that we can compare output after the fix. Differential Revision: https://phab.mercurial-scm.org/D9732

narrow: overwrite readfast in excludeddirmanifestctx The excludeddirmanifestctx does not have underlying storage (because it is excluded) so the implementation of the `readfast` trying to access it crash. This was reveled while running some cache warning code on narrow repositories. That code will be introduced further down in the series. Differential Revision: https://phab.mercurial-scm.org/D9731

patch: handle filenames with trailing spaces I have no idea if this is *actually* supported by the patch file format, but at least when reading from a patch file created by running `hg shelve`, it is written out such that there's a trailing space after the second (`b`) filename. When we read the patch file, we remove the space before parsing the filenames, so it doesn't end up matching the other sources of what files are in the shelve. We observed this internally due to a wrapper around unshelve that called into patch.changedfiles, but `hg patch` is able to reproduce the issue as well, so I've included both tests. Differential Revision: https://phab.mercurial-scm.org/D9729

rust: fix testing with $TMPDIR ≠ /tmp Differential Revision: https://phab.mercurial-scm.org/D9670

ci: avoid a global before_script definition This removes the implicit assumption that all jobs are test runs. Differential Revision: https://phab.mercurial-scm.org/D9723

setup: when possible, build and bundle man pages This makes it so the manual pages are built as part of the Python build, and includes them in any wheel generated. This should make Python wheels a much more useful and complete way of distributing Mercurial binaries. Differential Revision: https://phab.mercurial-scm.org/D9640

scmutil: improve documentation of writereporequirements() This makes it easier to understand the difference between `writerequires()` and `writereporequirements()`. Differential Revision: https://phab.mercurial-scm.org/D9568

command: automatically create alias for command using '-' in names As discussed during the "5.6 Sprint" ignoring dash in command name open the way to using them more. We can now move existing command to more readable (dash using) names without breaking any compatibility. Differential Revision: https://phab.mercurial-scm.org/D9515

worker: POSIX only supports workers from main thread (issue6460) The POSIX backend sets signal handlers for SIGINT (maybe avoidable) and SIGCHLD (necessary for waitpid). Python up to 3.9 only allow this from the main thread, so disable the worker feature otherwise. Differential Revision: https://phab.mercurial-scm.org/D9660

sharesafe: introduce config to disallow outdated shares if upgrade fails After this patch, we have config option to control all aspects of shares when share source is upgraded or downgraded. Differential Revision: https://phab.mercurial-scm.org/D9692

sharesafe: make warning about outdated share configurable If the source repository upgrades to use sharesafe mode, we show a warning in shares. This patch makes that warning configurable and some might not want their users see this warning. Differential Revision: https://phab.mercurial-scm.org/D9691

debuglock: rename flag names to better clarity `--force-lock` sounds as if we are taking the lock however in reality it's the opposite. Differential Revision: https://phab.mercurial-scm.org/D9690

sharesafe: add functionality to automatically downgrade shares Reasoning is same as previous patch which adds automatic upgrade support. Downgrade is required as if automatic upgrade is enabled, all shares upgrade and then source repository downgrades, shares won't work. We need to downgrade them. Differential Revision: https://phab.mercurial-scm.org/D9680

sharesafe: introduce functionality to automatically upgrade shares In past few months, we have developed a `share-safe` mode for sharing repository in which share source requirements and config values are shared with the shares. To get it rolling, an important task is to get these shares automatically upgraded. We are focusing on an installation where shares are created by scripts and test jobs. It will be difficult to manually upgrade these and we need some functionality to do so automatically. This patch introduces a config option to deal with it. If all of the following conditions are met, we upgrade the share repository automatically: * If the config option is enabled * Share source repository is share-safe enabled * Share is not share-safe enabled * Any command is run in the share Upgrading the share is pretty easy as it involves only editing the requirements file. Differential Revision: https://phab.mercurial-scm.org/D9679

localrepo: move storevfs calculation out of if statement In next patch, we will need this variable in else statement too. So, let's take it out. Differential Revision: https://phab.mercurial-scm.org/D9681

setup: don't import distutils prior to checking FORCE_SETUPTOOLS I've seen warnings about importing distutils before setuptools, although I can't reproduce them at the moment. Differential Revision: https://phab.mercurial-scm.org/D9722

largefiles: remove unused imports This fixes test-check-pyflakes.t Differential Revision: https://phab.mercurial-scm.org/D9747

upgrade: don't perform anything if nothing to do Before this patch, upgrade will process everything, re-clone all revlogs, write requirements file again even there is no change to be made. This patch makes it exit earlier. Differential Revision: https://phab.mercurial-scm.org/D9695

downgrade: if a compression is removed, consider that too For compression format variant, the result of `fromrepo()` and `fromconfig()` is not a boolean and we have to actually equate those to find change. Differential Revision: https://phab.mercurial-scm.org/D9693

engine: prevent a function call for each store file Python function calls are not cheap. Instead of calling the function once for each file in store, we use a dedicated function which filters and yields the required files. Differential Revision: https://phab.mercurial-scm.org/D9673

rhg: use a release-mode executable in tests This allows the rhg build for test-rhg.t to share compiled dependencies such as hg-core with the hg-cpython build for other tests. For context, my wrapper script for the typical edit-compile-test cycle now looks like this: (cd rust && cargo +nightly-2020-10-04 fmt) (cd rust && cargo build --release -p rhg) make --silent local PURE=--rust python test/run-tests.py --local "$@" Differential Revision: https://phab.mercurial-scm.org/D9728

rust: fix file folding map The file folding map, frequently used on macOS, had two issues: * the means for converting it to Python didn't work * a minor typo when copying the python code, where `!=` became `==` With this, the rust code passes all tests on macOS. Test Plan: I'm currently doing a full test run on a case-insensitive file system. If it passes, perhaps we can change the platform check from an error to a warning? Differential Revision: https://phab.mercurial-scm.org/D9671

test: replace a many occurence of `python` with `$PYTHON` Otherwise this can use the wrong python version, or worse, not find any python at all. Differential Revision: https://phab.mercurial-scm.org/D9730

tests: make test-subrepo-git.t compatible with git's master->main rename Git is about to change the default branch from "master" to "main". Recent versions has started warning about that, which makes tests fail. I assume the tests would fail in a different way once the default has changed. To make us compatible with that name change, let's set configure the default to be "master". That value makes the tests still work on older Git version (those where the default branch name is not configurable). Differential Revision: https://phab.mercurial-scm.org/D9746

revlog: use size_t for nodetree capacity This allows handling revlog containing more than 33554432 (INT_MAX / sizeof(nodetreenode)) revisions on x64 platforms. Differential Revision: https://phab.mercurial-scm.org/D9745

upgrade: demonstrate that a no-op upgrade still performs everything An upgrade operation which is not adding or removing anything should ideally error out. However, it does the whole cloning and everything. Next patch will fix it. Differential Revision: https://phab.mercurial-scm.org/D9694

upgrade: migrated -> upgraded in ui messages Differential Revision: https://phab.mercurial-scm.org/D9676

upgrade: remove unnecessary `is None` check `upgrade_engine.upgrade()` always return the `backuppath` value and there are not early returns. Hence I don't see how `backuppath` can be None. Adding extra unncessary safe checks hides unknown bugs. Hence removing it. Differential Revision: https://phab.mercurial-scm.org/D9675

engine: refactor code to replace stores in separate function In not all upgrades, we need to change the whole store. For example, when upgrading repository to share-safe mode, we don't touch revlogs at all hence store cloning and copying is not required. The store replacing code needs to be made aware about what all has changed and hence only copy/rename those things. To kickstart that, this patch moves existing logic into a separate function. Differential Revision: https://phab.mercurial-scm.org/D9674

engine: make hook point for extension a public function Well there are no private public functions in Python, but we generally treat functions and variables starting with `_` as private ones. A function which needs to be overrided in extension should be a public one. Differential Revision: https://phab.mercurial-scm.org/D9672