pathauditor: make _checkfs_exists a static method This fixes the bug detected by pytype where the auditor used in vfs.py may be a no-op auditor (vfs.py, line 398), which doesn't have the _checkfs_exists method.

merge: add mergeresult.mapaction to improve speed As a part of [hg update] we convert all [ACTION_CREATED] merge results into [ACTION_GET] actions, and that's slightly inefficient because every insertion pays the full cost of maintaining the [mergeresult] data structure up to date. This commit adds a function [mapaction], which is faster. (saves around 0.3s on a large update involving ~400k files)

merge: avoid dereferencing repo fields repeatedly

merge: skip syntactic path checks in [_checkunknownfile] We don't need to check the paths syntactically, since they are coming from diffing the revisions, so hopefully already checked on the way in. We still need to check what's on the filesystem, to avoid traversing the symlinks or subdirs, which we can't know about statically. Also, we use the directory audit to elide [isfileorlink], this removing ~all lstat calls from hg updates from-empty.

pathutil: use `finddirs_rev_noroot` instead of `parts` The benefit this brings is very tiny, if it's even there, since we still didn't get rid of the [parts] computation. It probably won't be worth it without the subsequent patch that adds one more use of [finddirs_rev_noroot]

pathutil: add the more efficient finddir iterator (to be used in subsequent commits)

typing: import unconditionally This is needed if we're writing python3 signatures.

tests: check how hgweb handles HEAD requests This test file is loosely based on test-hgweb.t. HEAD support originally implemented in fda5a4b853ab.

tests: test hg status --all with hgext/git There's a TODO item in git/dirstate.py about obtaining clean files in a more straightforward way, let's first test that status can and does show clean files at all.

tests: aborting on unknown revision emits exit code 10 even with hgext/git

worker: avoid reading 1 byte at a time from the OS pipe Apparently `pickle.load` does a lot of small reads, many of them literally 1-byte, so it benefits greatly from buffering. This change enables the buffering, at the cost of more complicated interaction with the `selector` API. On one repository with ~400k files this reduces the time by about ~30s, from ~60 to ~30s. The difference is so large because the actual updating work is parallellized, while these small reads are bottlenecking the central hg process.

diffutil: rewrite an ersatz ternary operator for building diffopts.text Pytype tends to get confused by these, but the real problem here is that if the caller passes `opts={}` as TortoiseHg has been doing, that is set as the value for `diffopts.text` when in reality it expects a boolean. When `None` is passed explicitly, the default value in `mdiff.defaultopts` is assigned.

mdiff: add a __str__ method to diffopts This makes it easier to debug by just formatting the object into `%s` to see the members and state, instead of the class and memory address.

debugshell: allow TortoiseHg builds to exit with the usual `quit()` command I've long been annoyed that `quit()` only randomly worked to exit the interpreter. When that happens, Ctrl+C doesn't work either (it simply prints "KeyboardInterrupt"), so then you have to `import sys` and `sys.exit()`. But it turns out that the behavior isn't random and it depended on which `hg.exe` was picked up on PATH first, because py2exe disables site initialization. I wasn't able to persuade the maintainer to allow an opt-in to initialization[1], but this works around it so that the behavior is now consistent however `hg.exe` is built. TortoiseHg 6.3.3 will be the first build that includes the site package, so handle the ImportError. [1] https://github.com/py2exe/py2exe/issues/154

doc: add a few comments

rhg: implement checkexec to support weird filesystems In particular, some of our repos are stored on a fileserver that simulates POSIX permissions poorly, in such a way that prevents the removal of execute permission. This causes rhg show a spurious unclean status, even though python hg reports the repo as clean. We fix this by making rhg implement the ~same checkexec logic that python hg does.

typing: use python3-style type annotation

merge: cache the fs checks made during [_checkunknownfiles] this ~halves the number of lstat calls made when updating from rev(-1) to a revision with lots of files

merge: disable the whole filesystem access loop if [_realfs] is false This makes it clearer that [auditeddir] is only relevant for [_realfs] checkers, and makes the non-realfs checkers more performant.

merge: short-circuit the _checkfs loop upon getting ENOENT This reduces the number of [lstat] calls when updating from rev(-1) to a rev with lots of files by a factor of several: for path foo/bar/baz/quux.txt without this patch we're lstatting: foo foo/bar foo/bar/baz foo/bar/baz/quux.txt and with this patch: foo foo/bar/baz/quux.txt

merge: don't pay for pathconflicts if there are none

pathauditor: no need to normcase the paths The only thing normed paths are used is the key of the caching sets, so the only change of behavior will be that the checks will be repeated for paths that differ by case. If anything, it seems correct for the check to be repeated, in case that actually affects semantics, but the main reasoning is simplifying the code and making it a bit faster. It looks like the code originally comes from commit [081e795c60e0]: it looks like that commit tried to get rid of the existing norming, but presumably did this overly cautiously, preserving it for the cache keys, even though it was pointless even then.

pathutil: slightly faster path audit in the common case

debug: add a config to abort update early This is useful to benchmark the parts of [hg update] that come before the parallel worker.

dirstate: add missing methods and kwargs to idirstate interface

dirstate: swap pathto() and get_entry() in idirstate This way the order of methods in dirstate and idirstate classes is the same. Just to make it easier to use diff tools to compare the two classes.

dirstate: update docstrings in idirstate from the current dirstate

setup: drop legacy osx compiler tuning to enable universal builds This was triggering deprecation warnings about migrating to `packaging.version` from `distutils` Version classes with `make local`. But rather than migrate that code, let's just get rid of some ~10-12 year old workarounds. As a bonus, the cext libraries that are built are now universal binaries containing x86_64 and arm64 images (at least when built on macOS 11.4 with Xcode 12.5 and the universal version of Python 3.9.13). Several things to note here: - Apple dropped support for 10.15 in Nov 2022, and OS X Lion that is referenced is 10.7 (unsupported since late 2014) - `xcode4` was basically always True because of the `>=` check (10.8 used Xcode 5, and I have Xcode 10.2 on 10.14) - `xcode51` was always False for modern-ish Xcode, because of the exact version string matching - Python 3.8 only supports OS X 10.9+; the Python 3.9.1+ universal installer is macOS 11+ only, and Python 3.10 drops the x86_64 installer to deliver only the universal installer. All of this is to say, the only thing lost by dropping this code on modern Xcode is that `os.environ['ARCHFLAGS'] = ''` is no longer set. But we probably shouldn't be setting that anymore, as shown by the universal libraries now being generated. I was able to `make local` and `python3 run-tests.py --local` with python 3.9.9, Xcode 10.2, and macOS 10.14.6, and didn't incur any more than the usual few test errors, so this should still work on some older versions of macOS.

tests: simplify `(py3 no-py36 !)` output matching predicates to `(no-py36 !)` It's all py3 now.

tests: drop `(py3 !)` output matching predicates Presumably these were paired with `(no-py3 !)` at one point, but now they were unconditionally true. test-check-code.t required a couple of `(glob)` markers on the changed lines in test-lfs-serve-access.t, because of the `$LOCALIP` usage on those lines. Not sure how those lines slipped through the checks previously.

run-tests: drop a workaround for python2 The problem is only on python3, but the awkward handling was because python2 didn't have this exception type. I've sporadically seen it running in WSL, but no clue what it means.

contrib: drop py2 support from testparseutil.py