branching: merge stable into default This show that the recent changes on default fixed the issue with transaction overwriting content in `test-transaction-wc-rollback-race.t`

rhg: in path_encode, simplify a bit more Use the slices for `basename` and `ext` instead of dealing with offsets.

rhg: in path_encode, be a bit more conservative about memory usage Use [shrink_to_fit] to match the previous behavior more closely, and potentially save (a tiny bit) of memory. FWIW, I suspect this is unnecessary, but this whole MR is about simplifying things while preserving any existing optimizations.

rhg: small refactor: stop using a magical constant "+ 1" Instead, directly do what the "+ 1" was supposed to achive: call hash_encode.

rhg: in path_encode, use Vec directly instead of VecDest No need to have a trivial wrapper over the type. There's nothing confusing about vec.write_bytes(...), after all.

rhg: in path_encode, split Dest into VecDest and MeasureDest Two separate types make the write semantics easier to understand because we can consider the two sinks separately. Having two independent compiled functions for size measurement and for actual encoding seems likely to improve performance, too. (and maybe we should get rid of measurement altogether) Getting rid of [Dest] also removes the ugly option rewrapping code, which is good.

rhg: use generic DestArr in hash_mangle This simplifies code a bit more, but comes with an extra memory copy in case [destlen == dest_vec.len()]. This is probably fine, but a follow-up change is removing that too.

rhg: in path_encode, make DestArr generic over its size

rhg: in path_encode add a DestArr type This is an implementation of Sink trait that writes into a fixed-size buffer on the stack, so identical to what was done before, but it makes the code of [hash_encode] easier to understand by dropping all these slice manipulations.

rhg: reduce verbosity in path_encode by using a trait for writing Hopefully this makes the code easier to read and understand and shorter overall. It also lets us later tweak the type we use as a [Sink], without having to change the encoding functions, including using two different types for size measurement and for the actual serialization.

refactor: simplify code in rust version of path_encode Moving the addition of '/' separator to the end of the loop makes the rest of the logic much simpler because the first iteration is no longer special.

dirstate: phase-divergent update to 4e95341c89aa Heptapod published the obsolete version of those.

dirstate: phase-divergent update to 65943224c184 Heptapod published the obsolete version of those.

dirstate: phase-divergent update to 3433723d1b9b Heptapod published the obsolete version of those.

dirstate: enforce change context for hacky_extension_update_file This was the last method not scoped yet

large-files: use a `changing_files` context when initializing the dirstate We are obviously mutating the dirstate, so lets scope this mutation.

dirstate: enforce context set_clean and set_possibly_dirty We don't want them called within a `changing_parents` context, but we still want them called within a context. So we update the decorator accordingly

keyword: wrap dirstate mutation in `changing_files` context This is the way.

keyword: wrap dirstate mutation in `changing_files` context This is the way.

dirstate: enforce `running_status` context for calling `status` Now that the context is working as intended and that the callers are updated. We can enforce it.

dirstate: have `running_status` warn when exiting with a dirty dirstate If running_status was started without the lock, all changes should have been explicitly written (with the lock) or invalidated before exiting the context.

dirstate: have `running_status` write the dirstate when holding the lock This is simple and harmless.

dirstate: check that dirstate is clean at the initial context opening More checking that we are not doing anything weird.

dirstate: start tracking that we are within a `running_status` context

dirstate: add documentation about the expectation of `running_status` context

contrib-perf: use `running_status` in `perf::status` This is the way.

large-files: also open the context in the subdirstate

large-files: use `running_status` in `mergeupdate`

large-files: use `running_status` in `scmutiladdremove` This is the way.

large-files: open the transaction sooner in `scmutiladdremove` We want it to encompass the status call.

large-files: use `running_status` in `overriderevert` This is the way

large-files: use `running_status` in `updatestandinsbymatch` This is the way.

large-files: wrap reposetup's status in a `running_status` context This is the way.

narrow: use `running_status` in `updateworkingcopy` This is the way.

status: use `running_status` in dirstate status This is the way.

status: pre-indent the dirstate status code This make the next changeset clearer.

dirstate: introduce a (noop) running_status context Let us start with a simplistic context so we can scope the appropriate code before adding more logic.

status: invalidate dirstate on LockError If we cannot take the lock, someone else is modifying the repository. Let us discard dirstate uncommitted data before exiting the status code. Having a clean dirstate after such operation seems safer. Strictly speaking, there is a small behavior change in the following situation: * process A call `status` outside of the `wlock` * process B grab the `wlock` * process A fails to acquires the lock to write status fixup * process B release the `wlock` *without touching the dirstate* * process A later grab the `wlock` * process A can write dirstate update from earlier `status` However this is a fairly hypothetical situation : * process A has to be raced * process B have to not update the dirstate * process A has to run another *unrelated* operation later. This seems rare enough to overlook. I am stating that the two operations in process A has to be unrelated. Otherwise, collecting status data outside of the lock to use them inside the lock is racy. Any other process could move things around (eg: the working copy) making the data collected during status irrelevantor even harmful. If such code exists, it should be fixed ASAP.

status: simplify the post status fixup phases With the wlock automatically discarding changes when applicable, we can simplify the code a bit. * we perform the fixup operation before trying to grab the lock to narrow the `try/except` * we no longer need to explicitly complare dirstate identities. We can trust the dirstate internal refresh for that. It would invalidate dirty data when needed. * detect still data invalidation by checking the dirty flag before and after taking the lock. Doing this is actually only necessary to issue the debug message, we could blindy trust the dirstate internal to ignore the `write` call on a non-dirty dirstate.

dirstate: cleanup the `_map` property cache The removed code was duplicating the effect of `@propertycache`. This is a gratuitous cleanup.

dirstate: only reload the dirstate when it may have changed This reinstall the equivalent of what the `filecache` was doing. However it does it at the dirstate level. There is a double motivation for this: - This avoid duplicating logic with the dirstate "identity" logic. - This increase the lifetime of the `dirstate` object, helping to implement change scoping.

dirstate: directly manage the dirstate property on localrepo Before we had: * the filecache layer on `localrepo` doing some caching * the dirstate having some internal invalidation/refresh machanism * the status code doing some identity validation. To clean this up, we are dropping the first item "localrepo `filecache`" from the equation in a favor of an approach integrated into the dirstate (second item) in the next changesets. This changeset will be a small windows where some things will be a bit slower. This will be fixed in the next changesets.

dirstate: factor the identity getting/setting code in the dirstate map We are doing the same things twice and we will add more logic in the next changesets. So lets start factoring things out now.

dirstate: use `cachestat` object for dirstatemap identity There is a class dedicated to this kind of cache check, let us use it. We will generalize this code in the next changesets, but we do the "behavior changing" pass on our own.

automv: lock the repository before searching for renames I detected this while debugging something else.

dirstate: distinct transaction callback from largefile This has not caused any issue so far because the large-files dirstate bypass the transaction logic. We might want to change that. In anyway, let us disarm this bug nest before it actually explode.

dirstate: track that changes are pending in a transaction Nothing is currently broken because if this, but this make the `_invalidated_context` attribute more accurate. Being more accurate here will help us later, when dealing with `status` call.

dirstate: add small asserts for double security We don't need this, but it does not hurt.

dirstate: simplify the invalidation management on context exit Since the `invalidate` call will directly reset the `_invalidated_context` attribut, we can simplify the code. In the same go, we move most of the logic out of the `finally` clause. It seems cleaner and safer. If we are handling an exception, we don't need the `write` code anyway.

dirstate: use the new `check_invalidated` decorator for `_changing` WeeeEEeee, less code.

dirstate: introduce a check_invalidated decorator This is a common need, so let us consolidate it to simplify the code.

dirstate: warn if dirty when starting an edition The dirstate should be clean before we start changing it. Otherwise we might write unrelated changes. Having a dirty dirstate laying around is also suspicious. This is similar to what we do when opening a new transaction, but this time this affect dirstate changes outside of a transaction.

large-files: make sure we write newly initialized standin file early Any changing context will have to initialized it before anything else. Not flushing the default (pre-change) content mean we would enter the changing context with a dirty dirstate, which is odd.

dirstate: mark `clear` and `rebuild` as `require_changing_parents` Yeah, more scoping!

dirstate: add a comment about the semantic of `dirstate.clear` This method is weird, lets flag it as such.

debugrebuildstate: wrap the operation in a `changing_parents` context This ismaybe a "changing_files" case? However this would be the only usage of `rebuild` outside a `changing_parents` context and this is a debug command, so lets not make the code base more complex because of that one command.

strip: use a `changing_parents` context for --keep update These are now properly scoped. note: it would be neat to reuse this in `hg rollback`.

mq: wrap the dirstate's rebuild in a `changing_parents` context This code is dealing with `qreshesh` failure. In that case the working copy will be left on the parent of the refreshed patch, so the parents are changing and `changing_parents` make sens.

lfconvert: use a `changing_parents` context to clear the dirstate Not sure if this is the right context, but it works and it is consistent with the other usages of `dirstate.clear`.

dirstate: mark the `copy` method as requiring a `changing_any` context This is used both when changing parents (e.g. merging with rename) and changing files (e.g. running `hg rename`).

dirstate: add a `require_changing_any` decorator We will need it for a couple of usecase (e.g `dirstate.copy`).

rebase: scope parent change into a changing_parents context If we are actually altering the working copy (i.e. we are not in memory), we should properly scope the working copy update.

dirstate: requires being in a `changing_parents` `context to set_parents` Enforcing proper operation scoping on all methods that mutate the dirstate will tighten correctness and reduce the risk of bugs. The context to use for this method is obvious, and all code was already compliant ☺

dirstate: invalidate on all exceptions Previously, we would miss SystemExit, KeyboardInterrupt etc. This "fix" on the bug tested in "test-largefiles-update.t" by preventing the precisely tested situation to happens at all. However this reveal a similar bug with a different timing. I have not been able to deal with that pre-existing bug so far. So I updated the test to point that out.

large-files: prepare a test for more changes The behavior around this test is about to change. We update the test to make it more robust and the changes clearer.

large-files: larger "changing_parents" context in mergeupdate override This since we are updating the lfdirstate early, it seems reasonable to include the full function in that scope.

large-files: use `hacky_extension_update_file` one more time This override is about merging and can be used in a `changing_parents` context. So lets use the method dedicated to hacky stuff when doing hacky stuff.

typing: disable `signature-mismatch` warnings on a few bytestr functions Recent versions of pytype complain about this, but it seems like expected behavior.

branching: merge with stable

rhg: fix a bug in path_encode This makes rhg able to access long paths at the root of the tree just fine.

rhg: demonstrate a bug in path_encode The bug means that long filenames at the root of the tree are encoded incorrectly by rhg, so rhg crashes when trying to access them.

rhg: nicer error message

typing: add type hints to argument checking functions in cmdutil These might be surprising, since they can take strings instead of bytes. The way `AnyStr` works is that it must be all bytes or all str for any given invocation. The wildcard here will be the `opts` that get passed in- if the type is unknown and defaults to `Any`, there's no enforcement that the dict key type matches the additional args. But a lot of uses should be using `**opts` from the command method, which has a str key. The uses of these methods in this module are now typed because their internals force a specific type, and it can't just be inferred from the caller.

tag: move the prohibition of tagging the `null` rev up to the `wdir()` check It makes sense to do these together, and avoid another revision lookup.

branching: merge with default the first part of the fix is there

tests: remove unnecessary --traceback argument

dirstate: remove the dedicated backup logic When alone, the dirstate can now take care of its commit/rollback pattern itself. When in a transaction, the transaction deal with commit/rollback pattern quite fine. Why did you have a dedicated backup logic?

localrepo: stop doing special dirstate backup at transaction open Since the dirstate writes are already managed by the transaction, we already do a backup of the dirstate when necessary (and even trigger one to keep `hg rollback` happy). We needs some special code to deal with the initial empty checkout, but it is not too complicated. Managing variable filename (as dirstate-v2 uses) at the "journalfile" level, is complex and fragile (which is consistent with the fact these files are not journal…). If we no longer do it, our life is significantly simpler. In some sense, we apply the xkcd-1134¹ solution to our savebackup/restorebackup problem. [1] https://xkcd.com/1134/ (the change to test-hardlink are expect as decreasing the number of duplicated backup drive the hardlink count down)

localrepo: "blindly" do a dirstate backup at the end of the transaction Having the file backup mechanism dealing with file backup as benefit. So lets move closer to that. The fact `hg rollback` even needs this is sad. I hope to have the time to implement one of the alternative soon.

test-hardlink: stop explicitly listing `undo.dirstate` The files is on its way out.

localrepo: enforce a clean dirstate when the transaction open This is important to simplify the backup process. This also seems like a quite reasonable expectation. See inline documentation for details.

dirstate: explicitly backup the datafile Since the transaction does not know about this file, we need to explicitly mark it for backup before touching it.

mq: write the dirstate before stripping I am not certain why it is dirty (probably the `repo.setparent` call a right before. However I know we need it flushed before the repository starts to be stripped.

dirstate: simplify the shelve hack to not go through the disk We already have the data in memory, so why not simply keep the data in memory? This avoid abusing the `savebackup/restorebackup` logic and will make our life easier.

dirstate: detect potential fishy transaction patterns while changing If we rely on the transaction more, we should be more careful about the transaction. Adding extra security seems reasonable.

dirstate: generalize the dirstate's invalidation on transaction abort The previous code was too specific, we should add the invalidation at next to the `addfilegenerator` in the `write` call.

dirstate: simplify some methods' decorator Since `changing_parents` and `changing_files` are mutually exclusive, having both `@requires_changing_files` and `@requires_not_changing_parents` on a method is redundant.

dirstate: document the functions that need consolidation They are more functions that make the dirstate dirty but does not currently enforce a clean change scoping, we add comments in front of each of them so clarify this. There is no urgency to it, but the world will be a better place when all of them have proper scoping in place.

dirstate: make `restorebackup` more robust when it is a noop If there are no data_file, we might not be able to determine its filename. Which can be safely ignored. (all this code is on the way out anyway)

dirstate-guard: remove the feature The dirstate guard duplicated some of the logic already implemented in the transaction (and now the changing_* context). However the feature was incomplete, for example, living only in memory meant we could not recover from the hardest crash. In addition this duplicated with the transaction logic meant things could go out of sync or step on each other. Removing the feature now that we no longer needs it seems the safest.

rollback: remove the dirstateguard usage Thanks to the previous changeset, we no longer needs it. begone !

rollback: explicitly skip dirstate rollback when applicable instead of letting the transaction logic overwrite the dirstate and then overwrite that overwrite with a backup. We simply do not restore the dirstate related file during the _playback call. This open the way to removing the last dirstate guard usage.

rollback: detect "parentgone" case earlier Detecting this earlier will help us to affect the rollback process sooner, which will help the next changesets. To keep things simple, we degrade the behavior a bit when the `undo.desc` is missing. However since `hg rollback` is not supposed to be used and the `undo.desc` file have been introduced in mercurial 1.6 (2010). I think this is an acceptable evil.

rollback: avoid a `hg commit --addremove` at a critical point The rollback behavior around `hg commit --addremove` has changed slightly. It does not really matters here but keeping that variant out of the way cannot hurt.

rollback: display some graphlog before/after a test piece This make the situation clearer.

rollback: show that the safety works in a associated test That test section is checking that --force is overcoming the safety check, however we did not check that the safety properly detect the situation. We now do for clarity.

dirstate-guard: remove its usage in `backout` We can simply replace it with a transaction.

dirstate-guard: remove the usage in `import` It is now redundant with the transaction spawning the same scope.

dirstate-guard: replace a usage in `rebase` with a transaction Opening the transaction sooner will provide us with the same benefit.

dirstate-guard: remove usage in `rebase` Now that the dirstate change and write are clearer, it does not seems we need to use a dirstate-guard here anymore. The transaction already wrap the full operation.

dirstate-guard: remove it usage in `mq` The code it covered is also covered by a `changing_parents` context.

dirstate: enforce the use of `changing_files` context to change tracking Now that everything is using the new context, we can start enforcing its usage.

dirstate: warn if we write to the dirstate without holding the wlock Writing the dirstate without holding the wlock can race with other update and overwrite important change. The comment questionning the current state of things was right, we should the `wlock` should always cover the dirstate. (Yes, this is me, agreeing with myself, half a decade later).

dirstate: avoid transaction backup/restore if we do not hold the lock Doing overwriting the dirstate content without holding the lock is a recipe for disaster.

dirstate: issue a developer warning on implicit write on wlock release Our goal is to get rid of all these to clarify the writing pattern, so it is time to warn about this (and later, forbid it).

status: fix post status invalidation If the dirstate changed under us, we should throw away what we have a reload it, should we not ?

status: fix post status writing With dirstate-v2, the status process itself might update internal states. So we make sure this get written on disk

dirstate: use `dirstate.change_files` to scope the change in `shelve` This is the way.

dirstate: use `dirstate.change_files` to scope the change in `unshelve` This is the way.

shelve: adjust what happens in some `changing_parents` context It seems like the rest is more about changing tracked_files, so we let start with moving them out of the `changing_parents` contexts.

dirstate: use `dirstate.change_files` to scope the change in `lfconvert` This is the way.

largefiles: rely on main scoping for writing dirstate in `markcommitted` Yeah, cleaner code.

largefiles: rely on main scoping for writing dirstate in `mergeupdate` Yeah, cleaner code.

largefiles: rely on the higher level `changing_giles` in `mergerecordupdates` Now that context open on the main dirstate also affect the underlying one, we can skip opening our own in `mergerecordupdates`

dirstate: use wlock and `dirstate.change_files` to scope the change in `mq` This is the way.

subrepo: use `changing_files` context in subrepository code This is better, not ideal, but better.

subrepo: let black expand some call on multiple lines early newer version of black are righfully doing this, so I apply it before more semantic change to reduce noise in the next changeset.

dirstate: use `dirstate.change_files` to scope the change in `import` This is the way.

dirstate: use `dirstate.change_files` to scope the change in `automv` This is... mostly... the way.

dirstate: use `dirstate.change_files` to scope the change in `amend` This is the way.