obsstore: fast path to check if obsstore is empty

obsstore: delay loading markers from obsstore file This will allow us to use cached revisions without parsing obsstore. Because _version isn't determined at __init__, the debugobsconvert command no longer works correctly. I'll fix it by a separate patch for the evolve extension.

obsstore: initialize _all markers without using _addmarkers() The next patch will make _all variable propertycached to avoid costly parsing of obsstore. This means we can't call _addmarkers() to initialize _all. Because all cached markers depend on _all, it isn't necessary to update these caches when _all is initially loaded.

revset: avoid implicit None testing in revset Implicit None testing is a very good way to get in trouble. We explicitly test for None.

revset: speed up existence checks for ordered filtered sets Previously, calling 'if foo:' on a ordered filtered set would start iterating in whatever the current direction was and return if a value was available. If the current direction was ascending, but the set had a fastdesc available, this meant we did a lot more work than necessary. If this was applied without my previous max/min fixes, it would improve max() performance (this was my first attempt at fixing the issue). Since those previous fixes went in though, this doesn't have a visible benefit in the benchmarks, but it does seem clearly better than it was before so I think it should still go in.

revset: remove existence check from min() and max() min() and max() would first do an existence check. Unfortunately existence checks can be slow in certain situations (like if the smartset is a list, and quickly iterable in both ascending and descending directions, then doing an existence check will start from the bottom, even if you want to check the max()). The fix is to not do the check, and just handle the error if it happens. In a large repo, this speeds up: hg log -r 'max(parents(. + .^) - (. + .^) & ::master)' from 3.5s to 0.85s. That revset is contrived and just for testing. In our real case we used 'bundle()' in place of '. + .^' Interesting perf numbers for the revset benchmarks: max(draft() and ::tip) => 0.027s to 0.0005s max(author(lmoscovicz)) => 2.48s to 0.57s min doesn't show any perf changes, but changing it as well will prevent a perf regression in my next patch. Result from revset benchmark revset #0: draft() and ::tip min max 0) 0.001971 0.001991 1) 0.001965 0.000428 21% revset #1: ::tip and draft() min max 0) 0.002017 0.001912 1) 0.001896 94% 0.000421 22% revset #2: author(lmoscovicz) min max 0) 1.049033 1.358913 1) 1.042508 0.319824 23% revset #3: author(lmoscovicz) or author(mpm) min max 0) 1.042512 1.367432 1) 1.019750 0.327750 23% revset #4: author(mpm) or author(lmoscovicz) min max 0) 1.050135 0.324924 1) 1.070698 0.319913 revset #5: roots((tip~100::) - (tip~100::tip)) min max 0) 0.000671 0.001018 1) 0.000605 90% 0.000946 92% revset #6: roots((0::) - (0::tip)) min max 0) 0.149714 0.152369 1) 0.098677 65% 0.100374 65% revset #7: (20000::) - (20000) min max 0) 0.051019 0.042747 1) 0.035586 69% 0.016267 38%

update: move default destination into a revset This is another step toward having "default" destination more clear and unified. Not all the logic is there because some bookmark related computation happened elsewhere. It will be moved later. The function is private because as for the other ones, cleanup is needed before we can proceed.

merge: move default destination computation in a revset This is another step toward having "default" destination more clear and unified.

mergecmd: simplify conditional The previous if tested the same thing with 'if repo._activebookmark'. We make the if/else logic clearer before a bigger refactoring.

rebase: move destination computation in a revset This is the first step toward making the "default destination" logic more clear and unified. The revset is private because I'm happy to delay the bikeshedding until after the clean up happened.

debian: include bash completions in mercurial-common

debian: install hgk as part of mercurial-common (issue4829) Leaving the hgk binary in /usr/bin causes some lintian warnings, and downstream packages poke it in /usr/share/mercurial, so we'll just stash it in there. Rather than patch hgk.py as part of the Mercurial install, just drop a config file in /etc/mercurial/hgrc.d that points to the installed hgk.

debian: install config files as part of mercurial-common Future patches will start putting config files in /etc/mercurial/, this just installs them.

debian: configuration so that cacerts is properly loaded This will be included in the mercurial-common package in a followup patch.

debian: install hg-ssh to /usr/bin just like downstream

highlight: remove temporary hack to overwrite repo.getcwd Now hgweb does it globally.

hgweb: overwrite cwd to resolve file patterns relative to repo (issue4568) It's useless to handle file patterns as relative to the cwd of the server process. The only sensible way in hgweb is to resolve paths relative to the repository root. It seems dirstate.getcwd() isn't used to get a real file path, so this patch won't cause problem.

dirstate: state that getcwd() shouldn't be used to get real file path hgweb will force it to be '' so that file patterns can be resolved relative to the repository root. I want to clarify that is correct.

merge: move merge step to the end Resolving other conflicts before merge ones is better because the state before the merge is as consistent as possible. It will also help with future work involving automatic resolution of merge conflicts with an external merge driver. There are no ordering issues here because it is easy to verify that the same file is never in both the dg/dm and the m sets.

lock: factor out lock testing into a separate function This is going to be needed for upcoming work with lock inheritance.

lock: factor code to read lock into a separate function This is going to be needed for upcoming work with lock inheritance.

tests: add unit tests for locking code We're going to make significant changes to lock behavior soon.

gitweb: port code selection without line numbers from paper This is adapted from f2e4fdb3dd27 and e92d4b8530cb. It also fixes issue4790 in gitweb; tab characters now have meaningful width on the modified pages (file view, file diff, changeset).

obsolete: remove unused _knownrevs function The call site was removed at cd62532c62a1.

bookmark: do not crash when active bookmark is forward and --date is used The logic that decides where to update according to the active bookmark location (when not on ".") was setting the rev to update to before we process --date. This lead to --date processing aborting because of duplicated specification. We reorder the two pieces of code and add a test for this.

bookmark: actually test update behavior in both cases The test claimed to test pull --update behaved the same way as pull + update, but the behavior of pull + update was never tested. We now test both.

coal: reuse style-paper.css and use a separate css file for visual tweaks To remove another big chunk of duplication from coal hgweb style, we can move all visual tweaks to style-extra-coal.css and include it together with style-paper.css.

coal: reuse all templates from paper/map by %including the file Before, coal/map simply had all the templates copied and pasted from paper/map file. It was easy to make a change to paper that would get into coal only partially and just sit there until something like 399e970e35c8 happened. Let's remove this duplication from coal/map by using %include to reuse templates from paper/map directly.

help/config: reword committemplate external editor text

help/config: reword committemplate multibyte character text

help/config: use :hg:`help config....` notation

help/config: omit True or False

help/config: capitalize items

tests: add more doctests for check-translation deprecated

tests: cleanup check-translation deprecated

gitweb: add author to filelogentry The line is copied from shortlogentry template.

monoblue: add author to filelogentry The line is copied from shortlogentry template.

monoblue: split a long line in filelogentry template The whole td has a nowrap class, which disallows breaking the line, so it's not required to use non-breakable spaces.

readbundle: fix typo of None compression The test simulates pre-d1b16a746db6 hgweb that sends "unbundle" capability with no argument.

readbundle: map 'HG10UN' to None compression In line with the other previous changes

getsubset: use None to request uncompressed changegroup

bundle: use None when unpacking uncompressed changegroup Let's be more modern!

writebundle: use 'None' instead of 'UN' for the bundle2 case Let's be modern!

compression: use 'None' for no-compression This seems more idiomatic and clearer. We still support both None and 'UN' for now because no user are migrated.

changegroup: move all compressions utilities in util We'll reuse the compression for other things (next target bundle2), so let's make it more accessible and organised.

obsolete: clarify that 'successorssets' returns the latest successors We do not return the first successors we found, we returns the latest (non obsolete (mostly)) one following the obsolete link transitively. We update the documentation to make this clean.

obsolete: drop an empty line at the end of a docstring This empty line is useless, let's drop it.

traceback: allow providing a local support contact point The extensions blaming code is fine for casual users but pretty terrible for corporate environments that can deploy a large amount of extensions to unsuspecting users. Reports will likely blame a random "innocent" extension (in our case crecord) and the hint in the message will triggers endless debug attempts from the user. We introduce a "ui.supportcontact" option that allow such big company to redirect their users to their own support desk. This disables all extensions blaming and just point people to the local support in all cases.

dispatch: pre-indent the extensions blaming logic We are about to put some real conditional here, but pre-indenting in it's own changesets makes next patch much cleaner.

tests: check for inconsistently translated DEPRECATED Mercurial expects DEPRECATED to be translated consistently, not doing that breaks Mercurial.

i18n-fr: reject translation which is missing DEPRECATED

i18n-pt_BR: select a single definition of DEPRECATED Mercurial expects DEPRECATED to be translated consistently, not doing that breaks Mercurial.

i18n-sv: select a single definition of DEPRECATED Mercurial expects DEPRECATED to be translated consistently, not doing that breaks Mercurial.

i18n-ro: use ()s instead of [] for DEPRECATED

i18n-ro: include translation of DEPRECATED

i18n-zh_CN: include translation of DEPRECATED

i18n-it: include translation of DEPRECATED

i18n-da: include translation of DEPRECATED

test-i18n: always run non gettext tests

localrepo: refresh filecache stats only if transaction finished successfully If commit is aborted by pretxncommit hook, in-memory changelog and manifest have entries that would be added. So they must be discarded on invalidate(). But the mechanism introduced by a710936c3037 doesn't handle this case well. It tries to mitigate the penalty of invalidate() by marking in-memory cache as "clean" on unlock assuming that they are identical to the stored data. But this assumption is wrong if stored data are rolled back by tr.abort(). This patch moves the hook to post-close action so that it will never be triggered on abort. This bug was originally reported to thg, which is only reproducible in command-server process on unix, evolve disabled. https://bitbucket.org/tortoisehg/thg/issues/4285/

localrepo: move closure of lock release to class It only captures "self", so it isn't necessary to be created dynamically.

highlight: add highlightfiles config option which takes a fileset (issue3005) Highlight extension lacked a way to limit files by size, by extension, and/or by any other part of file path. A good solution would be to use a fileset, since it can check file path, extension and size (and more) in one expression. So this change introduces such an option, highlighfiles, which takes a fileset and on each request decides if the requested file should be highlighted. The default "size('<5M')" is, in a way, suggested in issue3005. checkfctx() limits the amount of work to just one file (subset kwarg in fileset.matchctx()). Monkey-patching works around issue4568, otherwise using filesets here while running hgweb in directory mode would say, for example, "Abort: **.py not under root", but this fix is very local and probably far from ideal. I suspect there to be a way to fix this for the whole hgweb and resolve the issue, but I don't know how to do it.

posix: shellquote do not require quoting for "+" (issue4818)

hgweb: consume generator inside context manager (issue4756) If code inside a context manager returns a generator, the context manager exits before the generator is iterated. hgweb was using a context manager to control thread safe access to a localrepository instance. But it was returning a generator, so there was a race condition between a previous request streaming a response to the client and a new request obtaining the released but in use repository. By iterating the generator inside the context manager, we ensure we don't release the repo instance until after the response has finished. With this change, hgweb finally appears to have full localrepository isolation between threads. I can no longer reproduce the 2 exceptions reported in issue4756. test-hgweb-non-interactive.t has been modified to consume the output of calling into a WSGI application. Without this, execution of the WSGI application stalls because of the added yield statement.