sshpeer: break "OutOfBandError" feature for ssh (BC) When we'll be using the ssh's 'stderr' for realtime output successfully, it will no longer be possible to use 'stderr' to carry the error message (because it is already consumed by the real time output logic. This feature is very rarely used (test says largefile only) and I think breaking its output pattern is worth the benefit of having real time output with ssh.

error: allow a 'hint' to OutOfBandError This will be useful when changing the behavior of OutOfBandError for ssh in the next changeset.

tests: fix test-clone on vfat

wireproto: remove unused 'store' import

match: allow unioning arbitrary match functions A future patch will be allowing nested matchers. To support that, let's refactor _buildmatch to build a list of matchers then return True if any match. We were already doing that for filesets + regex patterns, but this way will be more generic.

match: add root to _buildmatch A future patch will make _buildmatch able to expand relative include patterns. Doing so will require knowing the root of the repo, so let's go ahead and pass it in.

localrepo: extract stream clone application into reusable function The existing stream_in method assumes a streaming clone is applied via the wire protocol. Previous patches have enabled streaming clone data to be produced and consumed outside the context of the wire protocol. However, the consuming part was incomplete because it didn't deal with things like updating the branch caches or writing out a requirements file. This patch finishes the separation of stream clone handling from the wire protocol. After this patch, it is possible to consume stream clones from arbitrary sources, including files. Mozilla plans to leverage this to serve pre-generated stream clone files to consumers, drastically reducing the wall and CPU time required to clone large repositories. This will enable clones to be nearly as fast as `tar`.

exchange: move code for consuming streaming clone into exchange For reasons outlined in the previous commit, we want to make the code for consuming "stream bundles" reusable. This patch extracts the code into a standalone function.

exchange: move code for generating a streaming clone into exchange Streaming clones are fast because they are essentially tar files. On mozilla-central, a streaming clone only consumes ~55s CPU time on clients as opposed to ~340s CPU time for a regular clone or gzip bundle unbundle. Mozilla is deploying static file "lookaside" support to our Mercurial server. Static bundles are pre-generated and uploaded to S3. When a clone is performed, the static file is fetched, applied, and then an incremental pull is performed. Unfortunately, on an ideal network connection this still takes as much wall and CPU time as a regular clone (although it does save significant server resources). We like the client-side wall time wins of streaming clones. But we want to leverage S3-based pre-generated files for serving the bulk of clone data. This patch moves the code for producing a "stream bundle" into its own standalone function, away from the wire protocol. This will enable stream bundle files to be produced outside the context of the wire protocol. A bikeshed on whether exchange is the best module for this function might be warranted. I selected exchange instead of changegroup because "stream bundles" aren't changegroups (yet).

dirstate: avoid match.files() in walk()

match: introduce boolean prefix() method tl;dr: This is another step towards a (previously unstated) goal of eliminating match.files() in conditions. There are four types of matchers: * always: Matches everything, checked with always(), files() is empty * exact: Matches exact set of files, checked with isexact(), files() contains the files to match * patterns: Matches more complex patterns, checked with anypats(), files() contains roots of the matched patterns * prefix: Matches simple 'path:' patterns as prefixes ('foo' matches both 'foo' and 'foo/bar'), no single method to check, files() contains the prefixes to match For completeness, it would be nice to have a method for checking for the "prefix" type of matcher as well, so let's add that, making it return True simply when none of the others do. The larger goal here is to eliminate uses of match.files() in conditions (i.e. bool(match.files())). The reason for this is that there are scenarios when you would like to create a "prefix" matcher that happens to match no files. One example is for 'hg files -I foo bar'. The narrowmatcher also restricts the set of files given and it would not surprise me if have bugs caused by that already. Note that 'if m.files() and not m.anypats()' and similar is sometimes used to catch the "exact" and "prefix" cases above.

hgweb: descend empty directories in monoblue The ability to "skip" a chain of empty directories in hgweb was added in c21d236ca897, but monoblue style wasn't updated. This block is copied from gitweb/map file and just works.

match: have visitdir() consider includes and excludes match.visitdir() used to only look at the match's primary pattern roots to decide if a treemanifest traverser should descend into a particular directory. This change logically makes visitdir also consider the match's include and exclude pattern roots (if applicable) to make this decision. This is especially important for situations like using narrowhg with multiple treemanifest revlogs.

hgweb: remove artificial width constraint from header in monoblue This width property comes from the beginning of the monoblue theme itself, and was used to stop the action header ("summary", "shortlog", "changelog", etc) from clashing with the search form. But it still was happening (on smaller screens, and with more actions added to hgweb over time). Effectively, the hardcoded width was preventing the header from fitting into the available screen space, since it always tried to be 900px wide, even if that meant horizontal scroll on smaller screens and having the actions on two lines where one should've been enough. For example, http://selenic.com/hg/log?style=monoblue has the last two actions ("gz" and "help") in the header on the second line, even when there seems to be enough space on the first. This patch makes the form float, which prevents it from overlaying/clashing with the action header, and allows the latter to resize itself in the best possible way.

merge with stable

files: recurse into subrepos automatically with an explicit path

dirstate: use open/read of vfs(opener) explicitly instead of read This simplifies changes in subsequent patch, which tries to open `.pending` file when HG_PENDING environment variable is defined.

dirstate: use self._filename instead of immediate string `dirstate` This prevents immediate string `dirstate` from multiplying. This is also a preparation for making dirstate aware of PENDING mechanism in subsequent patches.

revset: drop translation marker from error message of _notpublic() It is a kind of an internal error. End user won't see it.

revset: drop docstring from internal _notpublic() function It shouldn't be listed in "hg help revset".

record: make hg record always use the non curses interface Before this patch, hg record was running hg commit -i, therefore with the experimental.crecord=True flag, hg record was actually launching the curses record interface. Some of our users could be confused by that. This patch makes the hg record command set this flag to False, ensuring that hg record never shows the curses interface. commit -i, shelve -i and revert -i remain unchanged and use the curses interface if the experimental.crecord flag is set.

treemanifest: lazily load manifests Most operations on treemanifests already visit only relevant submanifests. Notable examples include __getitem__, __contains__, walk/matches with matcher, diff. By making submanifests lazily loaded, we speed up all these operations. The lazy loading is achieved by adding a _load() method that gets defined where we currently eagerly parse the manifest. We make sure to call it before any access to _dirs, _files or _flags. Some timings on the Mozilla repo (with flat manifest timings for reference): hg cat -r . README.txt: 1.644s -> 0.096s (0.255s) hg diff -r .^ -r . : 1.746s -> 0.137s (0.431s) hg files -r . python : 1.508s -> 0.146s (0.335s) hg files -r . : 2.125s -> 2.203s (0.712s)

treemanifest: speed up commit using dirty flag We currently avoid saving a treemanifest revision if it's the same as one of it's parents. This is checked by comparing the generated text for all three versions. Let's avoid that when possible by comparing the nodeids for clean (not dirty) nodes. On the Mozilla repo, this speeds up commit from 2.836s to 2.343s.

treemanifest: speed up diff by keeping track of dirty nodes Since tree manifests have a nodeid per directory, we can avoid diffing entire directories if they have the same nodeid. The comparison is only valid for unmodified treemanifest instances, of course, so we need to keep track of which have been modified. Therefore, let's add a dirty flag to treemanifest indicating whether its nodeid can be trusted. We set it when _files or _dirs is modified, and make diff(), and its cousin filesnotin(), not descend into subdirectories that are the same on both sides. On the Mozilla repo, this speeds up 'hg diff -r .^ -r .' from 1.990s to 1.762s. The improvement will be much larger when we start lazily loading subdirectory manifests.

localrepo: use correct argument name for pretxnclose hooks (BC) Before this patch, "the reason for the transaction" is passed to `pretxnclose` hooks via wrong name argument `xnname` (`HG_XNNAME` for external hooks)

localrepo: rename hook argument from TXNID to txnid (BC) From the first (3.4 or d283517b260b), `TXNID` is passed to Python hooks without lowering its name, but it is wrong.

test-walk: add more tests for -I/-X We had very limited testing of -I and -X, especially when combined with plain file patterns and with each other. This adds some more protection against regressions as upcoming patches modify the matcher code. (Originally meant for my own upcoming patches, but now I know drgott will be sending some patches soon.) The only noteworthy cases seems to be that both of hg debugwalk -Xbeans/black beans/black hg debugwalk -Xbeans beans/black walk the file. I would personally have expected the -X to trump. I don't care enough to change it, but I also think it's fair if some future commit changes the behavior.

ignore: use 'include:' rules instead of custom syntax Now that the matcher supports 'include:' rules, let's change the dirstate.ignore creation to just create a matcher with a bunch of includes. This allows us to completely delete ignore.py. I moved some of the syntax documentation over to readpatternfile in match.py so we don't lose it.

match: add 'include:' syntax This allows the matcher to understand 'include:path/to/file' style rules. The files support the standard hgignore syntax and any rules read from the file are included in the matcher without regard to the files location in the repository (i.e. if the included file is in somedir/otherdir, all of it's rules will still apply to the entire repository).

match: add optional warn argument Occasionally the matcher will want to print warning messages instead of throwing exceptions (like if it encounters a bad syntax parameter when parsing files). Let's add an optional warn argument that can provide this. The next patch will actually use this argument.

match: add source to kindpats list Future patches will be adding the ability to recursively include pattern files in a match rule expression. Part of that behavior will require tracking which file each pattern came from so we can report errors correctly. Let's add a 'source' arg to the kindpats list to track this. Initially it will only be populated by listfile rules.

check-code: reintroduce str.format() ban for 3.x porting In their infinite wisdom, the Python maintainers stripped bytes of its % and format() methods for 3.x. They've now added % back to 3.5, but format() is still missing. Since we don't have any particular need for it, we should keep avoiding it.

util: drop the 'unpacker' helper It is not helping anything anymore.

MBTextWrapper: drop dedicated __init__ method It was only there as a compatibility layer with a version of Python which we do support anymore.

util: drop the compatibility with Python 2.4 unpacker Python 2.4 compatibility have packed and sailed.

tests: just use 'response.reason' There is no reason to not have simple code now that Python 2.4 is dead.

url: drop awful hack around bug in Python 2.4 It's all just a memory now.

httpconnection: drop Python 2.4 specify hack Python 2.4.1 doesn't provide the full URI, good for it.

mail: drop explicit mail import required by Python 2.4 He's dead, Jim.

windows: drop Python2.4 specific hack for directory not found handling A good Python 2.4 hack is a removed Python 2.4 hack.

notify: drop import required by Python 2.4 Toto, I've a feeling we're not in anno 2004 anymore.

patchbomb: stop explicit import required by Python 2.4 Ding Dong, the witch is dead!

pager: drop python 2.4 hack around subprocess Farewell, we do not need you anymore.

check-code: drop ban of 'val if cond else otherval' construct We now have access to this horrible but less bad than 'cond and val or otherval' syntax.

check-code: entirely drop the 'non-py24.py' file from the test There are no Python 2.4 related errors remaining.

check-code: drop the 'format' built-in I'm not clear what it is doing, but one who knows what it is about can now make use of it.

check-code: drop ban of str.format After discussion with Augie and Matt, we are fine with it being introduced in the code base.

statichttprepo: remove wrong getattr ladder At least as far back as Python 2.6 the .code attribute is always defined, and to the best of my detective skills a .getcode() method has never been a thing.

merge with stable

match: explicitly naming a subrepo implies always() for the submatcher The files command supports naming directories to limit the output to children of that directory, and it also supports -S to force recursion into a subrepo. But previously, using -S and naming a subrepo caused nothing to be output. The reason was narrowmatcher() strips the current subrepo path off of each file, which would leave an empty list if only the subrepo was named. When matching on workingctx, dirstate.matches() would see the submatcher is not always(), so it returned the list of files in dmap for each file in the matcher- namely, an empty list. If a directory in a subrepo was named, the output was as expected, so this was inconsistent. The 'not anypats()' check is enforced by an existing test around line 140: $ hg remove -I 're:.*.txt' sub1 Without the check, this removed all of the files in the subrepo.

context: don't complain about a matcher's subrepo paths in changectx.walk() Previously, the first added test printed the following: $ hg files -S -r '.^' sub1/sub2/folder sub1/sub2/folder: no such file in rev 9bb10eebee29 sub1/sub2/folder: no such file in rev 9bb10eebee29 sub1/sub2/folder/test.txt One warning occured each time a subrepo was crossed into. The second test ensures that the matcher copy stays in place. Without the copy, the bad() function becomes an increasingly longer chain, and no message would be printed out for a file missing in the subrepo because the predicate would match in one of the replaced methods. Manifest doesn't know anything about subrepos, so it needs help ignoring subrepos when complaining about bad files.

ssh: capture output with bundle2 again (issue4642) I just discovered that we are not displaying ssh server output in real time anymore. So we can just fall back to the bundle2 output capture for now. This fix the race condition issue we where seeing in tests. Re-instating real time output for ssh would fix the issue too but lets get the test to pass first.

revset: optimize not public revset This patvh speeds up the computation of the not public() changeset and incidentally speed up the computation of divergents() changeset on our big repo by 100x from 50% to 0.5% of the time spent in smartlog with evolve. In this patch we optimize not public() to _notpublic() (new revset) and use the work on phaseset (from the previous commit) to be able to compute _notpublic() quickly. We use a non-lazy approach making the assumption the number of notpublic change will not be in the order of magnitude of the repo size. Adopting a lazy approach gives a speedup of 5x (vs 100x) only due to the overhead of the code for lazy generation.

phases: add set per phase in C phase computation To speed up the computation of draft(), secret(), divergent(), obsolete() and unstable() we need to have a fast way of getting the list of revisions that are in draft(), secret() or the union of both: not public(). This patch extends the work on phase computation in C and make the phase computation code also return a list of set for each non public phase. Using these sets we can quickly obtain all the revisions of a given phase. We do not return a set for the public phase as we expect it to be roughly the size of the repo. Also, it can be computed easily by substracting the entries in the non public phases from all the revs in the repo.

match: rename _fmap to _fileroots for clarity fmap isn't a very descriptive name for the set of the match's files.

match: remove unnecessary optimization where visitdir() returns 'all' Match's visitdir() was prematurely optimized to return 'all' in some cases, so that the caller would not have to call it for directories within the current directory. This change makes the visitdir system less flexible for future changes, such as making visitdir consider the match's include and exclude patterns. As a demonstration of this optimization not actually improving performance, I ran 'hg files -r . media' on the Mozilla repository, stored as treemanifest revlogs. With best of ten tries, the command took 1.07s both with and without the optimization, even though the optimization reduced the calls from visitdir() from 987 to 51.

dispatch: add support for python-flamegraph[0] profiling This gives us nicer svg flame graphs for output, which can make understanding some types of performance problems significantly easier. 0: https://github.com/evanhempel/python-flamegraph/

extensions: document that `testedwith = 'internal'` is special Extension authors (notably at companies using hg) have been cargo-culting the `testedwith = 'internal'` bit from hg's own extensions, which then defeats our "file bugs over here" logic in dispatch. Let's be more aggressive about trying to give extension authors a hint about what testedwith should say.

treemanifest: cache directory logs and manifests Since manifests instances are cached on the manifest log instance, we can cache directory manifests by caching the directory manifest logs. The directory manifest log cache is a plain dict, so it never expires; we assume that we can keep all the directories in memory. The cache is kept on the root manifestlog, so access to directory manifest logs now has to go through the root manifest log. The caching will soon not be only an optimization. When we start lazily loading directory manifests, we need to make sure we don't create multiple instances of the log objects. The caching takes care of that problem.

hook: drop dedicated catch for 'KeyboardInterrupt' This is no longer under 'Exception' in Python 2.6.