imports: allow importing futures from concurrent Differential Revision: https://phab.mercurial-scm.org/D12278

setup: remove pygit2 Python 2 logic Differential Revision: https://phab.mercurial-scm.org/D12277

setup: inline now-constant list This varied when we supported Python 2. Differential Revision: https://phab.mercurial-scm.org/D12276

setup: unconditionally do this python 3 step Differential Revision: https://phab.mercurial-scm.org/D12275

setup: remove Python 2 support code for determining dylib suffix Differential Revision: https://phab.mercurial-scm.org/D12274

setup: inline os.fsdecode now that we're done with Python 2 Differential Revision: https://phab.mercurial-scm.org/D12273

setup: inline encoding constant that is only used once This was variable back when we supported Python 2. Differential Revision: https://phab.mercurial-scm.org/D12272

setup: remove printf trampoline Differential Revision: https://phab.mercurial-scm.org/D12271

setup: remove more Python 2 support code I'll inline print() etc in future patches. Differential Revision: https://phab.mercurial-scm.org/D12270

setup: remove ssl check that only matters on 2.7 Differential Revision: https://phab.mercurial-scm.org/D12269

setup: remove block that tries to help Python 2.6 users It's time to move on folks. Differential Revision: https://phab.mercurial-scm.org/D12268

setup: drop statement of support for Python before 3.5.3 🎉🎉 Differential Revision: https://phab.mercurial-scm.org/D12266

cext: remove inline rewriting of argv This only worked on Python 2. And since we dropped support for Python 2, we can drop support for this functionality. Differential Revision: https://phab.mercurial-scm.org/D12233

cext: remove PY23() Since we always run on Python 3 now, we no longer need this macro to support Python 2. We refactor all users to just use the 2nd argument. Differential Revision: https://phab.mercurial-scm.org/D12232

cext: unconditionalize PySlice_GetIndicesEx() We only support Python 3 now. Differential Revision: https://phab.mercurial-scm.org/D12231

cext: unconditionalize PYLONG_VALUE() We only support Python 3 now. Differential Revision: https://phab.mercurial-scm.org/D12230

cext: use PyLong symbols We no longer support Python 2. So we can unconditionally use the Python 3 symbol names. Differential Revision: https://phab.mercurial-scm.org/D12229

cext: use PyLong symbols We no longer need to support Python 2. So use the Python 3 symbol names directly. Differential Revision: https://phab.mercurial-scm.org/D12228

cext: remove Python 2 support We still alias the Python 2 symbols. This will be cleaned up in a separate commit. Differential Revision: https://phab.mercurial-scm.org/D12227

cext: remove Python 2 file handling code Differential Revision: https://phab.mercurial-scm.org/D12226

cext: remove Python 2 variant of listdir_slot() Differential Revision: https://phab.mercurial-scm.org/D12225

cext: remove some conditional preprocessor defines We may want to inline these defines. But for now, getting rid of the Python 2 support is a step forward. Differential Revision: https://phab.mercurial-scm.org/D12224

cext: unconditionally use PyLong_FromLong() We no longer support Python 2. Differential Revision: https://phab.mercurial-scm.org/D12223

cext: drop preprocessor PyInt aliases Now that we dropped support for Python 2 we can use the Python 3 native functions. Differential Revision: https://phab.mercurial-scm.org/D12222

cext: remove Python 2 module initializer functions We no longer need these since we dropped support for Python 2. Differential Revision: https://phab.mercurial-scm.org/D12221

rhg: simplify the handling of share-safe config mismatch Differential Revision: https://phab.mercurial-scm.org/D12213

branching: merge stable into default

heptapod-ci: remove Python 2 support (hurray!) 6.1 was the last release to support Python 2, let's lighten the CI. Differential Revision: https://phab.mercurial-scm.org/D12206

relnotes: add 6.1.3 and 6.1.4

Added signature for changeset 0cc5f74ff7f0

Added tag 6.1.4 for changeset 0cc5f74ff7f0

purge: prevent a silly crash with --confirm --files if --files is passed, there was no directory to checks and `msg` was undefined. This is now fixed and tested.

rust-status: don't trigger dirstate v1 rewrite when only v2 data is changed The assumption that we need to rewrite (or append to) the dirstate if the ignore pattern hash has changed or if any cached directory mtimes have changed is only valid when using dirstate-v2. In dirstate-v1, neither of these things are written to disk.

commit: allow to close branch when committing change over a closed head Otherwise, an explicit other commit become necessary, which seems both silly and verbose. This is useful when merging closed heads on the same branches, for example when merging multiple repositories together.

logcmdutil: use the same data as {file*} template keywords (issue6642) Since 0c72eddb4be5 template keywords that show files use a different source of data than ctx.p1().status(ctx). These two functions in logcmdutil also show file lists when needed (e.g. log with --debug flag), but previously they used the old way of just looking at status compared to p1 and it resulted in differences between e.g. hg log --debug and hg log -T '{file*}'. test-phases.t needs an adjustment because 7 is a merge commit of two topological branches and one of them introduces files C, D and E.

tests: show that hg log --debug output differs from {file*} template keywords hg log --debug -T xml doesn't differ, but let's test it because we can.

docker: avoid /tmp write access issues by fixing permissions

url: raise error if CONNECT request to proxy was unsuccessful The deleted code didn’t work on Python 3. On Python 2 (or Python 3 after adapting it), the function returned in the error case. The subsequent creation of SSL socket fails during handshake with a nonsense error. Instead, the user should get an error of what went wrong. I don’t see how the deleted code would be useful in the error case. The new code is also closer of what the standard library is doing nowadays that it has proxy support (which we don’t use in the moment). In the test, I use port 0 because all the HGPORTs were already taken. In practice, there should not be any server listening on port 0.

revset: fix the doc of "nodefromfile" This should maybe be called "nodesfromfile", but at least the documentation is correct (it was previously a copy past from follow).

py3: don’t subscript socket.error On Python 2, socket.error was subscriptable. On Python 3, socket.error is an alias to OSError and is not subscriptable. The except block passes the exception to self.send_error(). This fails on both Python 2 (if it was executed) and Python 3, as it expects a string. Getting the attribute .strerror works on Python 2 and Python 3, and has the same effect as the previous code on Python 2.

demandimport: eagerly load msvcrt module on PyPy

ci: drop the phabricator refresh step Now that phabricator is no longer in us, we should avoid this useless step to save time and simplify things.

hghave: make black version regex work with newer versions of black Black commit 117891878e5be4d6b771ae5de299e51b679cea27 (included in black >= 21.11b0) dropped the string "version " from the output of "black --version". To make the regex work with newer black versions, make matching of "version " optional.

rust: relax im-rc dependency to allow minor updates This "15.0.*" requirement came from 0d99778af68a and is now replaced with plain "15.0". AFAICS, it really should allow (but not necessarily require) im-rc 15.1 . Narrow requirement requirements with wildcard in the version is not used in other places.

Added signature for changeset 6b10151b9621

Added tag 6.1.3 for changeset 6b10151b9621

narrow_widen_acl: enforce narrowacl in narrow_widen (SEC) Reviewer note: this was sent by the author as a simple bugfix, but can be considered a security patch, since it allows users to access things outside of the ACL, hence the (SEC) prefix. However, this affects the `narrow` extention which is still marked as experimental and has relatively few users aside from large companies with their own security layers on top from what we can gather. We feel (Alphare: or at least, I feel) like pinging the packaging list is enough in this case.

chg: ignore already closed fds when cleaning up This should fix this error we see in the CI from time to time: ``` --- /tmp/mercurial-ci/tests/test-chg.t +++ /tmp/mercurial-ci/tests/test-chg.t.err @@ -187,6 +187,26 @@ $ chg bulkwrite --pager=on --color no --config ui.formatted=True paged! 'going to write massive data\n' killed! (?) + Traceback (most recent call last): + File "/tmp/hgtests._uvojvqb/install/lib/python/mercurial/commandserver.py", line 509, in _serverequest + sv.cleanup() + File "/tmp/hgtests._uvojvqb/install/lib/python/mercurial/chgserver.py", line 382, in cleanup + self._restoreio() + File "/tmp/hgtests._uvojvqb/install/lib/python/mercurial/chgserver.py", line 461, in _restoreio + os.close(fd) + OSError: [Errno 9] Bad file descriptor + Traceback (most recent call last): + File "/tmp/hgtests._uvojvqb/install/lib/python/mercurial/commandserver.py", line 693, in _acceptnewconnection + self._runworker(conn) + File "/tmp/hgtests._uvojvqb/install/lib/python/mercurial/commandserver.py", line 744, in _runworker + prereposetups=[self._reposetup], + File "/tmp/hgtests._uvojvqb/install/lib/python/mercurial/commandserver.py", line 509, in _serverequest + sv.cleanup() + File "/tmp/hgtests._uvojvqb/install/lib/python/mercurial/chgserver.py", line 382, in cleanup + self._restoreio() + File "/tmp/hgtests._uvojvqb/install/lib/python/mercurial/chgserver.py", line 461, in _restoreio + os.close(fd) + OSError: [Errno 9] Bad file descriptor [255] ```

tests: work around non-thread-safeness of sysconfig.get_config_var()

workflow: add a default template for Merge Request Introduce a first basic template to try the feature.

run-tests: prevent race-condition when picking a channel Before this, multiple jobs could search the list at the same time and pick the same free channel. We now project this search/assignment with a simple lock.

run-tests: send the test result after freeing the channel Sending the message about the test being "done" signals to the main thread that a new test can be started. Before this changeset, we sent this signal before freeing the channel, there is room for a race condition where a new test would search for a channel before the old test freed the one it used. This is an example of the failure it would produce: https://foss.heptapod.net/mercurial/mercurial-devel/-/jobs/552404

run-tests: also send a message in the Keyboard interrupt case The next patch will do something equivalent, so lets do the change in an independant changeset first in case we need to bisect something in the future.

rhg: correctly handle the case where diffs are encoded relative to nullrev returning a valid entry for nullrev fix chain that delta against nullrev.

test-revlog: test a repository that contains a diff against nullrev We are witnessing a crash in the rust code, so we lets make sure this case is tested.

worker: adapt _blockingreader to work around a python3.8.[0-1] bug (issue6444) Python 3.8.0 is the latest I can load on Ubuntu 18.04, and I regularly hit the TypeError because this function is missing. While it can be avoided by disabling worker usage via config option, that's a bit obscure. I'm limiting the function definition to the narrow range of affected pythons because there were other bugs in this area that were worked around, that I don't fully understand. See the bug report for discussions on why the narrow range, and related commits working around other bugs. Differential Revision: https://phab.mercurial-scm.org/D12627

doc: use an absolute path in sys.path The idea and rationale is similar to https://phab.mercurial-scm.org/D12599 (landed as 1b6e381521c5). Differential Revision: https://phab.mercurial-scm.org/D12622

check-py3-compat: use an absolute path in sys.path The idea and rationale is similar to https://phab.mercurial-scm.org/D12599 (landed as 1b6e381521c5). Differential Revision: https://phab.mercurial-scm.org/D12621

rhg: align the dirstate v2 writing algorithm with python Use the same algorithm of file append as python does, where we do a manual seek instead of relying on O_APPEND. (see the reasons in the inline comment)

test-dirstate: actually test the append code path in dirstate v2 Apparently it's not sufficient to modify a file to force the dirstate write-out, so the append code path was untested. By removing a file instead of changing we're forcing append to happen.

ci: do not trigger phabricator for merge-request The fast the phabricator steps has a `rules` entry makes it selected for the special `merge_requests` pipelines. The other ones behave as default and are not selected tot the mrege_request pipelines. This result in a second pipeline to be created, with only the phabricator pipeline in it. Which usually succeed fast (since there is nothing to do). This is harmful as this create a false sense of "the series is passing" and Gitlab will use this simplistic pipeline for validation. By explicitly preventing the pipeline to be created in the merge-request case, we prevent this situation to happens Note that the job will be dropped (alonside phabricator) in the next two weeks anyway.

amend: stop specifying matcher, get all copies in wctx When we're recreating the commit that we'll be committing, we don't want to filter our copy information based on just the *new* [versions of the] files we're amending. The test has an example of this case, but for clarity, the situation is: ``` $ hg cp src dst && hg commit <do some work> $ hg amend some_unrelated_file.txt $ hg status --copies A dst A some_unrelated_file.txt ``` What *should* happen is that `dst` should remain marked as a copy of `src`, but this did not previously happen. `matcher` here only includes the files that were specified on the commandline, so it only gets the copy information (if any, in this example there's not) for `some_unrelated_file.txt`. When it goes to apply the memctx to actually create the commit, the file copy information is incomplete and loses the information for the files that shouldn't have been affected at all by the amend. Differential Revision: https://phab.mercurial-scm.org/D12625

amend: add test showing poor behavior when copies are involved Differential Revision: https://phab.mercurial-scm.org/D12624

censor: fix [hg update] away from a revision with censored files Differential Revision: https://phab.mercurial-scm.org/D12604

censor: demonstrate a bug Differential Revision: https://phab.mercurial-scm.org/D12584

Added signature for changeset 0ddd5e1f5f67

Added tag 6.1.2 for changeset 0ddd5e1f5f67

ci: remove py2-rust support Nobody cares about this very narrow usecase, and py2 support is over by July 1st. This helps with the CI load, and removes some flakiness.

relnotes: add release notes for 6.1.2

docs: use proper rst markup for preformatted blocks The multiple lines were re-flowed to a single line, both in man page and html.

test-dirstate: print something when the check is skipped This makes a programming error obvious in cases when it should not be skipped Differential Revision: https://phab.mercurial-scm.org/D12602

test-dirstate: fix detection of Rust environment variable The Rust path never actually worked. This change also improves clarity of the comment. The next change will ensure we print something when this check fails. Differential Revision: https://phab.mercurial-scm.org/D12601

rust-dirstate-v2: fix the unused bytes counter when rewriting the dirstate As per the previous patch, the counter was incorrectly carried over from the old docket when it should be reset for a complete rewrite. Differential Revision: https://phab.mercurial-scm.org/D12594

rust-dirstate-v2: show `unused_bytes` counter is not reset on total rewrite This was picked up by @aalekseyev when doing unrelated debugging. The Rust implementation was never resetting this counter, so a brand new file would carry over the old counter. As I write this, my counter is a supposed 7389089 unused bytes for a total of 170978 bytes in the data file. Feel free to post your own high score. Differential Revision: https://phab.mercurial-scm.org/D12593

test-dirstate: use new `--docket` flag to get the data uuid This is more robust. Differential Revision: https://phab.mercurial-scm.org/D12592

debugcommands: remove `debugdirstateignorepatternhash` This is replaced by the more complete `--docket` option to `debugstate`. Differential Revision: https://phab.mercurial-scm.org/D12591

dirstate-v2: add flag to `debugstate` to print docket information This is useful information that we don't easily have access to currently, unless you speak fluent xxd. This replaces `debugdirstateignorepatternshash`, which I'll remove in the next changeset. Differential Revision: https://phab.mercurial-scm.org/D12590

doc: use an absolute path in sys.path to work around a python DLL loading bug This bug[1] was causing the zstd entry under "Available Compression Engines" to be omitted from the documentation for some versions of python3. I could upgrade, but past upgrades have broken venvs and clobbered installed packages. It's a trivial workaround, so there's no sense in leaving this subtle issue. It was flushed out by changing the module policy here from 'allow' to 'c', and seeing this error: ImportError: DLL load failed while importing parsers: The parameter is incorrect. [1] https://github.com/python/cpython/issues/87271 Differential Revision: https://phab.mercurial-scm.org/D12599

rhg: fix dirstate-v2 data file removal system In D12581 I introduced logic to remove the previous dirstate-v2 data file after a new one is created (and its corresponding docket), but the logic was flawed. I fixed it and made it simpler to understand by gather all logic in a single expression. Differential Revision: https://phab.mercurial-scm.org/D12586

rhg: use `Command::exec` instead of `Command::status` `rhg` is supposed to be a transparent executable, using a subprocess defeats that purpose. See inline comments for more details. This also introduces the `which` crate to check if the fallback executable actually exists to help debugging (plain `execve` doesn't give much information). The error code 253 is used to signify that the fallback is not found, but may mean in the future that it is otherwise invalid if we start being more specific. Differential Revision: https://phab.mercurial-scm.org/D12578

commit: don't use hard-coded `.hg/last-message.txt` path in error message Whoever wrote D8463 just used a hard-coded path to `.hg/last-message.text` instead of using the relative path that was already available in the `msgfn` variable (and used just a few lines up in related message). Let's fix that. Differential Revision: https://phab.mercurial-scm.org/D12585

rust-dirstate-v2: clean up previous data file after the docket is written This was overlooked before and caused many data files to keep living forever. We could potentially consider adding a random cleanup of them in case they slipped through one day. Differential Revision: https://phab.mercurial-scm.org/D12581

rust-dirstate-v2: save proper data size if no new data on append This is currently only triggered with the tests ran with `--rhg` without `--rust`, by "luck", there probably always was something to write, like an mtime when also using Rust extensions alongside `rhg`. Differential Revision: https://phab.mercurial-scm.org/D12580

procutil: avoid `+= None` when writing to full std{err,out} descriptor on py3 The write function returns `None` if there was no room to write the given data[1]. I don't like that this is effectively an infinite loop if there's never any progress emptying the underlying buffer, but we're no worse off than before, and it fixes random stacktrace popups seen in the py3 build of TortoiseHg. [1] https://docs.python.org/3/library/io.html#io.RawIOBase.write Differential Revision: https://phab.mercurial-scm.org/D12555

amend: don't remove unselected removals from memctx When there are removed files in the working copy and they are not selected to be amended into the parent, the `filectxfn` we create for the `memctx` would still return `None` before this patch. That's clearly incorrect; we should return the `filectx` from the unamended commit. Somehow it seems to not matter much except for the case with copies stored in changesets. Thanks to Kyle Lippincott for doing all the debugging and identifying the fix for this issue. Differential Revision: https://phab.mercurial-scm.org/D12573

tests: demonstrate crash on partial amend with copies in changesets See the fix in the next patch for explanation. Differential Revision: https://phab.mercurial-scm.org/D12572

rebase: while rewriting desc hashes, ignore ambiguous prefix "hashes" If a repo is sufficiently large, a six digit number "hash prefix" can somewhat easily reference an ambiguous hash prefix. Differential Revision: https://phab.mercurial-scm.org/D12552

tests: add test demonstrating issue with ambiguous has prefixes during rebase Differential Revision: https://phab.mercurial-scm.org/D12551

help: set the large-file-limit to 10MB This is a minor increase (5%) and makes the doc much clearer. Differential Revision: https://phab.mercurial-scm.org/D12484

help: clarify the unit of `ui.large-file-limit` config Its might be a bit confusing, especially since `large-file.min-size` uses MB. Differential Revision: https://phab.mercurial-scm.org/D12483

debuglock: ignore ENOENT error when unlocking This is consistent with the main `lock.release` code. Differential Revision: https://phab.mercurial-scm.org/D12481

run-tests: introduce "forward-slash" version of everything on windows This should be useful for some shell invocation. Differential Revision: https://phab.mercurial-scm.org/D12480

tests-racy-mutation: pass the editor through config instead of env variable On Windows msys seems to do awful mangling of the environment variable content that confuses everything to the death. Going through the config works fine, so we do that instead. Differential Revision: https://phab.mercurial-scm.org/D12479

Added signature for changeset 5bd6bcd31dd1

Added tag 6.1.1 for changeset 5bd6bcd31dd1

relnotes: add notes for 6.1.1 This also fixes the header for 6.1 from 6.1rc0

rust-hgpath: add `repr(transparent)` to `HgPath` It's been stabilized a long time ago, so let's not rely on an implementation detail now. Differential Revision: https://phab.mercurial-scm.org/D12433

rust-dirstatemap: correctly decrement the copies counter This was caught when writing unit tests for the `DirstateMap`. We were always setting `had_copy_source` to `false` since we erased the value just before. Differential Revision: https://phab.mercurial-scm.org/D12432

rust-dirstatemap: properly decrement counter for tracked descendants I found this bug when writing unit tests after the fact for the `DirstateMap`. We never decremented the tracked descendants counter since we were always resetting the node data before reading it. This also drops the use of `state`, in favor of the new API to get that information. Differential Revision: https://phab.mercurial-scm.org/D12431

rust-dirstate: panic if the DirstateMap counters go below 0 When modifying the API I hit some... interesting errors (trying to allocate 178GB of RAM, for example) because I failed to keep the counters correctly updated. This counter underflow is likely to happen when code is changed around and can have up to eat-your-dirstate level of consequences, which is not nice. The very small runtime cost of checking these counters should really not be an issue and will help us uncover bugs when/if they do appear in the future. Differential Revision: https://phab.mercurial-scm.org/D12430

rust: fix unsound `OwningDirstateMap` As per the previous patch, `OwningDirstateMap` is unsound. Self-referential structs are difficult to implement correctly in Rust since the compiler is free to move structs around as much as it wants to. They are also very rarely needed in practice, so the state-of-the-art on how they should be done within the Rust rules is still a bit new. The crate `ouroboros` is an attempt at providing a safe way (in the Rust sense) of declaring self-referential structs. It is getting a lot attention and was improved very quickly when soundness issues were found in the past: rather than relying on our own (limited) review circle, we might as well use the de-facto common crate to fix this problem. This will give us a much better chance of finding issues should any new ones be discovered as well as the benefit of fewer `unsafe` APIs of our own. I was starting to think about how I would present a safe API to the old struct but soon realized that the callback-based approach was already done in `ouroboros`, along with a lot more care towards refusing incorrect structs. In short: we don't return a mutable reference to the `DirstateMap` anymore, we expect users of its API to pass a `FnOnce` that takes the map as an argument. This allows our `OwningDirstateMap` to control the input and output lifetimes of the code that modifies it to prevent such issues. Changing to `ouroboros` meant changing every API with it, but it is relatively low churn in the end. It correctly identified the example buggy modification of `copy_map_insert` outlined in the previous patch as violating the borrow rules. Differential Revision: https://phab.mercurial-scm.org/D12429

rust: explain why the current `OwningDirstateMap` is unsound See inline comments. Differential Revision: https://phab.mercurial-scm.org/D12428

dispatch: fix silly blackbox entries when hg is interrupted When hg is interrupted, it creates ui.log like this: 1970/01/01 00:00:00 user @0000000000000000000000000000000000000000 (62488)> killed! exited 255 after 1.78 seconds This is due to a scoping problem: two different uses of the name "msg" collide. So rename one of them. Differential Revision: https://phab.mercurial-scm.org/D12427

precheck: fix false warning about content-divergence creation Before this patch, if we try to `hg prune` (without any successors) an already obsoleted cset which has at least one successor, it would false warn about new content-divergence. As we know, pruning cset without any successors can not create any divergence. Differential Revision: https://phab.mercurial-scm.org/D12002

streamclone: avoid some obscure error in a corner case I don't really know how, but I ran into this error: $ hg clone --stream ssh://user@dummy/empty-repo local-empty-repo streaming all changes abort: unable to apply stream clone: unsupported format: [255] I think you need an empty list of requirements for this to happen, which is weird, but an obscure error like this is not exactly helpful either. Since this is the result of an encoding bug anyway, just fix it. Differential Revision: https://phab.mercurial-scm.org/D12402

tags: fix typo in fast path detection of fnode resolution (issue6673) If I understand it, mctx.readfast() is unreliable here if p1/p2 .hgtags nodes differ, and tags on that branch would be randomly discarded depending on which parent were picked. The test case added by this patch would fail only on zstd-compressed repository. I didn't try hard to stabilize the failure case.

dirstate-cext: properly invalidate mtime and data in `set_untracked` This was forgotten about in the initial implementation and was revealed while adding the `dirstate-v2` variant of `test-issue660.t`. Neither the existing Python implementation nor the upcoming Rust implementation suffer from this bug since they respectively have `None` and `Option<T>` to represent the lack of information. Differential Revision: https://phab.mercurial-scm.org/D12414

pullbundle: fix file name in the help text It is pullbundles.manifest and not pullbundle.manifest. Differential Revision: https://phab.mercurial-scm.org/D12391

unamend: abort if commit was not created by `hg [un]amend` `hg unamend` can currently undo any kind of rewrite, as long as it has an obsmarker. However, that has quite unexpected results if you run it after e.g. `hg rebase` (expecting it to behave like a generic `hg undo` command), because it updates to the predecessor and leaves the old changes in the working copy. I think it's better to allow `hg unamend` only after `hg amend` (and after `hg unamend` because that's documented as being supported). Differential Revision: https://phab.mercurial-scm.org/D12390

test: use `wait-on-file` in `test-racy-mutations.t` The official utility scale its timeout with the run-tests.py one. So lets use it. Differential Revision: https://phab.mercurial-scm.org/D12382

amend: fix amend with copies in extras If copy information is stored only in the commit extras and not in filelogs, then they get lost on amend if the file wasn't also modified in the working copy. That's because we create `filectx` object from the old commit in those cases, and the `.copysource()` of such objects read only from the filelog. This patch fixes it by always creating a new `memfilectx` in these cases, passing the calculated copy information to it. Differential Revision: https://phab.mercurial-scm.org/D12387

tests: demonstrate that copy info in changeset gets lost on amend When copy information is stored in changesets, it gets lost on amend. We didn't notice that until now because our users at Google have the config set to `compatibility`, which means copy information is stored in both changeset and filelogs. Differential Revision: https://phab.mercurial-scm.org/D12386