dirstate: restore original estimation and update comment The former comment didn't reflect the content of the dirstate entries, the two nodes are a fixed header in the file and not per-entry. Split the documented entry into the path part and the fixed header. The heuristic itself is still valid, e.g. for the NetBSD src tree a maximum path size of 142 and an average of 49, resulting in 66 bytes per entry on average. Differential Revision: https://phab.mercurial-scm.org/D8850

merge stable in default

dirstate: revert change to Rust binding The Rust binding uses its own class and is not derived from the regular implementation. As such, it didn't get _nodelen. Differential Revision: https://phab.mercurial-scm.org/D8849

relnotes: copy "next" to "5.5" and clear "next" The same procedure as every year^Wcycle. Differential Revision: https://phab.mercurial-scm.org/D8839

commitctx: extract all the file preparation logic in a new function Before we actually start to create a new commit we have a large block of logic that do the necessary file and manifest commit and that determine which files are been affected by the commit (and how). This is a complex process on its own. It return a "simple" output that can be fed to the next step. The output itself is not that simple as we return a lot of individual items (files, added, removed, ...). My next step (and actual goal for this cleanup) will be to simplify the return by returning a richer object that will be more suited for the variation of data we want to store. After this changeset the `commitctx` is a collection of smaller function with limited scope. The largest one is still `_filecommit` without about 100 lines of code.

commitctx: gather more preparation code within the lock context This is a small change that exist mostly for clarification. I am about to move a large amount of code in its own function. having all that code next to each other will make the next changeset clearer.

commitctx: move a special case about files earlier Same logic as a changeset a bit earlier, the `writefilecopymeta` section is more a post processing details so we move the conditional about `files` value closer to the rest of the code computing `files` value.

commitctx: extract all the manual logic to process the files That branch of the if is significantly more complicated than the other two. Moving it to its own function make it simple to keep the scope limited and to read to the higher level function.

commitctx: treat `filesadded` more like `filesremoved` Accumulating the filename in a list will have a negligible cost and deal with the list of added files like the other ones will make is code cleaning simpler. The two variable with very close name is not great, but my plan is to split most of the code in a separated function which will make the "problem" go away by itself.

commitctx: move `writechangesetcopy` business at the end a code section This code is to handle a specific subcase so we move it a the end. This allow to gather the rest of the "core" code closer to the related logic.

commitctx: move copy meta config reading in a dedicated function The logic is non trivial, make it contained in a function is clearer. It also unlock easy re-use of that logic without having the pass the value around.

commitctx: no longer use the `writecopiesto` variable in the function The `writefilecopymeta` variable already carry the same information, so we can use `writefilecopymeta` in the one conditional where `writecopiesto` was used.

commitctx: extract the function that commit a new manifest The logic is large enough and isolated enough to be extracted, this reduce the size of the main function, making it simpler to follow.

commitctx: stop using weakref proxy for transaction This weakref proxy was introduced in 2007 by 30d4d8985dd8. If I understand it correctly, the logic at that time was relying on the transaction destructor, triggered at garbage collection time to rollback failed transaction. passing the object to sub function directly mean it would live in the function scope and be trapped in the traceback on exception, leading to the transaction rollback too late in some case. Modern transaction usage use explicit opening and closing of transaction and no longer rely on some internal reference counting details. So this weakref proxy is no longer necessary. Absolutely no test are affected when we drop it.

rebase: fix regression in file change detection introduced by 0ecb3b11fcad Before 0ecb3b11fcad, `wctx._compact()`, was called by `wctx.nofilechanges()` as a side effect. Later, it turned out that this side effect is needed to correctly detect which files changed. See https://phab.mercurial-scm.org/D8733#131949 for the history. The problem could also be triggered by running `tests/test-rebase-parameters.t` with `--extra-config-opt rebase.experimental.inmemory=1`. Differential Revision: https://phab.mercurial-scm.org/D8843

context: re-add `overlayworkingctx._compact()` removed in 6a5dcd754842 This partially backs out 6a5dcd754842. The method was and is unused, but a call to it is introduced in the next patch. Differential Revision: https://phab.mercurial-scm.org/D8842

push: test for checks preventing publishing obsolete changeset The main difference from the previous test is how the changeset was obsoleted. In this case it is an amend so publishing the orphan would also create phase divergence. This must not go unnoticed.

push: test for checks preventing publishing obsolete changeset This introduce a simple example, more are coming. See inline documentation for details.

push: another test for checks preventing pushing orphaness to a server In this one, orphan was create with and amend instead of a prune.

push: test the checks preventing pushing orphaness to a server This is introduce a simple example, more are coming. See inline documentation for details.

commitctx: document the None return for "touched" value

tests: add tests trying to use dir as style, as %include, and as __base__ Differential Revision: https://phab.mercurial-scm.org/D8800

tests: fix mistaken copy&paste from commit 4489e9a22763 Differential Revision: https://phab.mercurial-scm.org/D8802

manifest: kill one more instance of the old merge hash hack Differential Revision: https://phab.mercurial-scm.org/D8766

dirstate: isolate node len dependency for the pure version When switching to a 256bit hash function, this still needs adjustment, but concentrates the change in one place. Differential Revision: https://phab.mercurial-scm.org/D8815

fix: update documentation to reflect preference for --source over --rev I should have updated the documentation in 5205b46bd887 (fix: add a -s option to format a revision and its descendants, 2020-03-13) and/or a6ef1e8e2f6d (fix: mark -r as advanced, 2020-03-13)... Differential Revision: https://phab.mercurial-scm.org/D8808

tests: make check-py3-compat.py actually load the specified files correctly For most uses, this change is essentially a no-op, as this script is generally only run by test-check-py3-compat.t, which will already put `$TESTDIR/..` in `$PYTHONPATH`. When running outside of tests, however, `$PYTHONPATH` is likely not set, causing check-py3-compat.py to parse the file from the repo, but then import the installed version, and raise any errors about the installed version, not the one currently in the repo. Additionally, this helps users (like me) who have a strange set up where their home directory (and thus their hg repos) happen to be in a subdirectory of sys.prefix (which is /usr on my system). Since the '.' entry added to sys.path takes precedence over the absolute path of `$TESTDIR/..` in `$PYTHONPATH`, the path to the modules that it imports (and that show up in any stack trace) are *relative*, meaning that we don't detect them as starting with `sys.prefix`. Sample non-test invocation, and the difference this change makes (the path for 'error at <path>:<line>' is correct now):: Before: ``` $ python3 contrib/check-py3-compat.py mercurial/win*.py mercurial/win32.py: error importing: <ValueError> _type_ 'v' not supported (error at check-py3-compat.py:65) mercurial/windows.py: error importing: <ModuleNotFoundError> No module named 'msvcrt' (error at check-py3-compat.py:65) ``` After: ``` $ python3 contrib/check-py3-compat.py mercurial/win*.py mercurial/win32.py: error importing: <ValueError> _type_ 'v' not supported (error at win32.py:11) mercurial/windows.py: error importing: <ModuleNotFoundError> No module named 'msvcrt' (error at windows.py:12) ``` Differential Revision: https://phab.mercurial-scm.org/D8814

tests: virtualenv is only used on py2, rename and conditionalize If I have I have the Debian `python3-virtualenv` package installed on my machine, the import succeeds but then I receive an AttributeError because the package is essentially completely different between py2 and py3, and test-hghave fails. Differential Revision: https://phab.mercurial-scm.org/D8812

morestatus: mention --stop even if not using --verbose Differential Revision: https://phab.mercurial-scm.org/D8811

exchange: backout changeset c26335fa4225 Changeset c26335fa4225 has good intends but introduce significant behavior regressions for multiple important cases. In short there are many case where push would have caught instability creation/propagation that are no longer covered. These behavior have been covered for many years and even if some related case are not currently caught, the covered one should not be regressed. The next four changesets introduce tests for some of these cases. However we could produce many more tests cases since the area is wide and they are many possible combination. (And we should cover them when getting back to this issue) Since 5.5 is one week away, the most reasonable approach seems to back this out while we devise a new way to move forward that preserve the current behavior, catch more issues and also improves the situation that c26335fa4225 target. In addition to the behavior change, c26335fa4225 also introduced output changes. These output changes does not requires a backout per-se, but are part of the same changeset. However they come with a couple of issues that also requires attention: 1) the bulk of the error message have been shoehorned into a multiple line abort message. This seems quite different from what we usually do. The abort message should be a compact and efficient message, with extra details being issued as normal error output beforehand. (with --verbose/--quiet) support. 2) the current output is unbounded, so if there is many (tens, hundreds, thousands, …) of unstable/obsolete changeset involved in the push, the output can quickly become a scary and un-usuable wall of text. So we need some limitation here (same as we have with the remote head list that says A, B , C and # others).