match: delete unused root and cwd arguments to constructors (API) Most matchers no longer need the root and cwd arguments. patternmatcher and includematcher still need the root argument for subincludes. Differential Revision: https://phab.mercurial-scm.org/D5929

dirstate: call and cache os.getcwd() in constructor I'm about to make scmutil.matchfiles() not pass the root and cwd paths to match.exact(), since they no longer have any effect. That turned out to have the surprising effect of making some tests (test-rebase-scenario-global.t and test-removeemptydirs.t) crash when the working directory was removed. The problem was that my patch removed the call to repo.getcwd(), which caused the current working directory to not be cached in the dirstate as early as it was before. This patch fixes that by caching the current working directory in the dirstate constructor. Differential Revision: https://phab.mercurial-scm.org/D5928

cleanup: prefer matchmod.{always,never}() over accessing matchers directly We have these factory methods so we should use them. Differential Revision: https://phab.mercurial-scm.org/D5927

match: allow passing in badfn to always() and never() So that no callers are required to access the matchers themselves. Differential Revision: https://phab.mercurial-scm.org/D5926

py3: use raw strings and %d for formatting Before the string compares on Python 3 failed because we were comparing bytes to str. Using raw strings ensures we are always comparing str. While we're here, also use %d to format integers. Differential Revision: https://phab.mercurial-scm.org/D5925

diff: make --stat respect ui.relative-paths It would have been easy to make all diffs respect ui.relative-paths, but we don't want that since it makes the diff invalid. Perhaps it makes sense to do that with --noprefix since the point of that is to make paths that are easy to copy&paste, and the diff is already invalid anyway. But this patch just makes the --stat version respect the config option. The --stat view is not even close to a valid diff, so I think it makes sense to show the paths in more human-friendly form. Differential Revision: https://phab.mercurial-scm.org/D5896

match: delete unused argument "listsubrepos" from _buildmatch() Seems to have been unused since 9f9ffe5f687c (match: compose 'set:' pattern as matcher, 2018-06-10). Differential Revision: https://phab.mercurial-scm.org/D5924

match: delete unused rel() (API) Differential Revision: https://phab.mercurial-scm.org/D5923

largefiles: get cwd and relative paths from repo instead of matcher As far as I can tell, the matcher returned from matchandpats() got its cwd from the context object, which in turn got it from the repo, so this should be equivalent. Differential Revision: https://phab.mercurial-scm.org/D5922

scmutil: remove special handling of pats==("",) in matchandpats() (API) I don't know under what circumstances we used to pass that value, but we don't seem to do it now. Differential Revision: https://phab.mercurial-scm.org/D5921

scmutil: respect ui.relative-paths in default match.badfn We should probably be passing in a uipathfn here instead, so the caller can get consistent output (between match.badfn and whatever else it prints). I think we should ideally remove the badfn from the matcher completely, but that's a different story. This patch is at least not making it worse. Differential Revision: https://phab.mercurial-scm.org/D5916

cmdutil: migrate previously missed instances to uipathfn I seem to have missed these two instances in 15f63ac122ea (files: respect ui.relative-paths, 2019-01-29) and 7068c6b0114b (revert: respect ui.relative-paths, 2019-02-05). Differential Revision: https://phab.mercurial-scm.org/D5915

cat: respect ui.relative-paths for "skipping missing subrepository" Differential Revision: https://phab.mercurial-scm.org/D5914

addremove: use uipathfn instead of m.rel() for recorded similatity message When no path arguments are given to addremove, it generally prints absolute paths. However, before this patch, we would always print the "recording removal of foo as rename to bar (78% similar)" message with relative paths. Differential Revision: https://phab.mercurial-scm.org/D5913

forget: use relative paths for --interactive Seems more consistent. Differential Revision: https://phab.mercurial-scm.org/D5912

automv: respect ui.relative-paths Differential Revision: https://phab.mercurial-scm.org/D5910

debugwalk: avoid match.rel() and use repo.pathto() instead I'm about to delete match.rel() Differential Revision: https://phab.mercurial-scm.org/D5909

match: delete unused abs() (API) Differential Revision: https://phab.mercurial-scm.org/D5908

copy: respect ui.relative-paths in copy/rename Differential Revision: https://phab.mercurial-scm.org/D5907

match: delete unused uipath() and _uipathrelative (API) Differential Revision: https://phab.mercurial-scm.org/D5906

diff: respect ui.relative-paths for warning about path outside --root I set legacyrelativevalue=True rather than being based on what patterns the user passed (as the current match.uipath() does). I think it simply doesn't really matter for this message (there are not even any tests for it). Differential Revision: https://phab.mercurial-scm.org/D5905

largefiles: use uipathfn instead of match.{rel,uipath}() (API) All callers now pass in a uipathfn, so we can just use that instead of choosing the right match function. This also means that when we make add/remove respect ui.relative-paths, it will work for largefiles too. Differential Revision: https://phab.mercurial-scm.org/D5904

forget: pass around uipathfn and use instead of m.rel() (API) Differential Revision: https://phab.mercurial-scm.org/D5911

addremove: pass around uipathfn and use instead of m.uipath() (API) Differential Revision: https://phab.mercurial-scm.org/D5903

remove: pass around uipathfn and use instead of m.rel() (API) Same as previous commit, but now for remove. Differential Revision: https://phab.mercurial-scm.org/D5902

add: pass around uipathfn and use instead of m.rel() (API) For now, the uipathfn we pass around always prints relative paths just like before, so this should have no effect. Well, there's one little change: I also made the "skipping missing subrepository: %s\n" message relative. Differential Revision: https://phab.mercurial-scm.org/D5901

subrepo: (mostly) use relative path in "skipping missing subrepository" This is consistent with the other messages printed by these functions. Note that addremove is a little different and prints absolute (aka repo-relative) paths if no argument was given. Differential Revision: https://phab.mercurial-scm.org/D5900

subrepo: use relative path for "already tracked" message From 932de135041f (subrepo: warn when adding already tracked files in gitsubrepo, 2015-02-27): The file is printed with abs() to be consistent with how it is printed in workingctx, even though that is inconsistent with how added files are printed in verbose mode. However, a few year later, the same author wrote 7008f6819002 (context: name files relative to cwd in warning messages, 2017-07-11) and now it's inconsistent. This fixes that inconsistency. Differential Revision: https://phab.mercurial-scm.org/D5899

diff: make `hg diff --root=subrepo` work The root prefix is currently stripped before adding the subrepo prefix. It seems to me that if you run `hg diff --root=subrepo` and a subrepo/ path is visited (say "subrepo/foo"), then the path generated for the diff would initially be just "foo" and we then fail (if the developer warning is active) to strip the "subrepo/" from the path. It seems like we should first add the prefix in order to produce a path that's relative to the top-level repo, and then we can remove the root prefix from that. There are no tests for this, it seems, and I don't care enough to add one. Differential Revision: https://phab.mercurial-scm.org/D5895

patch: replace "prefix" and "relroot" arguments by "pathfn" (API) The two arguments serve a very similar purpose: "relroot" is stripped from the front of the path, and then "prefix" (a subrepo path) is added (also to the front). Passing in a function that does that is more generic and will make it easier to respect ui.relative-paths in later patches (don't worry, I'm not going to respect that option for regular patches, only for --stat). I'm deliberately not calling it "uipathfn", because it's generally for producing valid diffs (including when prefix is non-empty), so things like using backslash on Windows is not an option. Differential Revision: https://phab.mercurial-scm.org/D5894