worker: support more return types in posix worker This allows us to return things that aren't tuple(int, str) from worker functions. I wanted to use marshal instead of pickle, but it seems to read from the pipe in non-blocking mode, which means it stops before it sees the results. The windows worker already supports arbitrary return values without serialization, because it uses threads instead of subprocesses. Differential Revision: https://phab.mercurial-scm.org/D3845

debug: process --debug flag earlier This allow the verbosity level to be set correctly during extension initialization.

windows: fix incorrect detection of broken pipe when writing to pager Paging e.g. hg incoming on Windows and quitting the pager before the output is consumed will print 'abort: Invalid argument'. This is because the windows error 0xE8 (ERROR_NO_DATA) is mapped to EINVAL even though it is documented as 'The pipe is being closed'. Note that this fix assumes that Windows' last error code is still valid in the exception handler. It works correctly in all my tests. A simpler fix would be to just map EINVAL to EPIPE, like was done is flush previously, but that would be less precise. This error was not observed previously, when pager was an extension.

test: stop passing --quiet in a run dedicated to debug output The goal of the run is to display some debug output. Passing --quiet in this case is strange.

acl: add bookmarks support Originally submitted at https://www.mercurial-scm.org/pipermail/mercurial-devel/2016-March/080650.html as an RFC by timeless. .. feature:: The `acl` extension now has support for bookmarks as well as branches. Differential Revision: https://phab.mercurial-scm.org/D3750

repository: define manifest interfaces The long march towards declaring interfaces for repository primitives continues. This commit essentially defines interfaces based on the following types: * manifest.manifestdict -> imanifestdict * manifest.manifestlog -> imanifestlog * manifest.memmanifestctx -> imanifestrevisionwritable * manifest.manifestctx -> imanifestrevisionstored * manifest.memtreemanifestctx -> imanifestrevisionwritable * manifest.treemanifestctx -> imanifestrevisionstored * util.dirs -> idirs The interfaces are thoroughly documented. Their documentation is now better than the documentation in manifest.py in many cases. With the exception of util.dirs, classes have been annotated with their interfaces. (I didn't feel like util.dirs needed the proper interface treatment.) Tests have been added demonstrating that all classes and instances conform to their interfaces. This work was much easier than filelogs. That's because Durham did an excellent job formalizing the manifest API a while back. There are still some minor kludges with the interfaces that should probably be addressed. But the primary goal with interface declarations is getting something established. Once we have an interface, we can modify it later easily enough. Differential Revision: https://phab.mercurial-scm.org/D3869

narrowbundle2: when we handle a widen, mark the operation as unsafe We have to manually drive the context manager here since the logic is awkwardly split between a couple of places. Differential Revision: https://phab.mercurial-scm.org/D3719

narrow: mark the critical chunks of narrowing/widening as unsafe I'm _mostly_ sure these are the only unsafe chunks here. Differential Revision: https://phab.mercurial-scm.org/D3718

repair: mark the critical section of strip() as unsafe When our experimental nointerrupt feature is enabled, this will help prevent users from corrupting their repo during a strip. Differential Revision: https://phab.mercurial-scm.org/D3717

ui: add an uninterruptable context manager that can block SIGINT The blocking of SIGINT is not done by default, but my hope is that we will one day. This was inspired by Facebook's "nointerrupt" extension, which is a bit more heavy-handed than this (whole commands are treated as unsafe to interrupt). A future patch will enable this for varying bits of Mercurial that are performing unsafe operations. It's intentional that the KeyboardInterrupt is raised as the context manager exits: during the span of the context manager interrupting Mercurial could lead to data loss, but typically those spans are fairly narrow, so we can let the unsafe block complete and then terminate hg (which will leave the repo in a consistent state, even if it's not the user's desired state). .. api:: New context manager ``ui.uninterruptable()`` to mark portions of a command as potentially unsafe places to interrupt Mercurial with Control-C or similar. Differential Revision: https://phab.mercurial-scm.org/D3716

merge with stable

Added signature for changeset 0b63a6743010

Added tag 4.6.2 for changeset 0b63a6743010

revset: add partial support for ancestor(wdir()) It's easy, so let's make it happen. I'm not certain if 'wdir() &' should be required. ancestors(wdir()) works without it, but ancestor(wdir()) doesn't as of now. That's the issue of fullreposet.__contains__() vs __and__().

context: also accept diff option directly For now we accept both diff option and dictionary. This will let us upgrade internal users gradually before we drop the old API at the end of this series.

context: also take all other arguments than `patch.diff` accept This is needed to use `context.diff` as a full replacement of `patch.diff`

context: explicitly take diffopts in `context.diff` (API) To provide a proper replacement for the `patch.diff(…)` function, the `context.diff(…)` method needs to be able to take more parameters. To distinguish the diff options from the new other arguments, we upgrade the diff options to its own explicit argument.

rebase: check whether the rebasestate exists or not a bit early Converted the else part into `if True` because that part will soon be under a except part. Differential Revision: https://phab.mercurial-scm.org/D3876

rebase: use staeobj to check whether interrupted rebase exists This cleans up the code a bit. Differential Revision: https://phab.mercurial-scm.org/D3875

rebase: add a stateobj variable to rebaseruntime class The stateobj variable is an object of state.cmdstate() and will be used in upcoming patches to clean up the logic a bit and also use cbor format to write data to rebasestate class. Differential Revision: https://phab.mercurial-scm.org/D3874

rebase: refactor logic to read rebasestate in a separate function This will help us in plugging the use of state.cmdstate() to read rebasestate. Differential Revision: https://phab.mercurial-scm.org/D3873

revlog: reuse 'descendant' implemention in 'isancestor' The two functions do the same thing, but one takes nodes while the other takes revs. Using one to implement the other make sense. We should probably cleanup the API at some point to avoid having so many similar functions. However, we focus on an efficient implementation for now.

revlog: efficient implementation of 'descendant' Iterating over descendants is costly, because there are no "parent -> children" pointers. Walking the other way around is much more efficient, especially on large repositories, where descendant walks can cost seconds. And the other hand, common ancestors code follows links in the right direction and has a compiled implementation. In real life usage, this saved up to 80s during some pull operations, where descendant test happens in extension code.

revlog: refactor out the rev-oriented part of commonancestorheads We plan to use this in a function taking revs as argument. Round trips to nodes seem silly.

revlog: do inclusive descendant testing (API) In many other places, a revision is considered a descendant of itself. We update the behavior of `revlog.descendant()` to match this. (for example. `revlog.isancestor` does inclusive testing). No tests break, so it seems safe to do so. This will make it easier to use a more efficient implementation in a later changeset.

manifest: make cachesize a private attribute AFAICT this isn't accessed outside the class. It is a private attribute and its naming should reflect that. Differential Revision: https://phab.mercurial-scm.org/D3868

repository: document that file() return value conforms to interface Differential Revision: https://phab.mercurial-scm.org/D3867

cleanup: pass in overwrite flag to hg.updaterepo() as named argument For clarity.

histedit: use self.stateobj to check whether interrupted histedit exists self.stateobj is an instance of state.cmdstate() class which has an .exists() function which is used to check whether there exists an interrupted statefile or not. Differential Revision: https://phab.mercurial-scm.org/D3865

histedit: add a stateobj variable to histeditstate class The stateobj variable will be an instance of state.cmdstate() class. The stateobj variable will be used in upcoming patches to simplify the code a bit and start using cbor to write state files. Differential Revision: https://phab.mercurial-scm.org/D3864

histedit: factor out logic of processing state data in separate fn The new function will serve as the point from where we always get a dictionary of data stored in the statefile and will be helpful in integrating state.cmdstate.read() to write statefile in cbor. Differential Revision: https://phab.mercurial-scm.org/D3863

histedit: use hg.updaterepo() to avoid ui.{push|pop}buffer() hack In some parts of code, we call hg.update() and surround that with ui.pushbuffer() and ui.popbuffer() to suppress the output returned by hg.update(). We have hg.updaterepo() which does not writes to UI and can be used instead. Differential Revision: https://phab.mercurial-scm.org/D3862

scmutil: fix __enter__ in progress context manager Differential Revision: https://phab.mercurial-scm.org/D3861

httppeer: fix use of uninitialized variable with devel logging If the request fails, "res" was uninitialized. Differential Revision: https://phab.mercurial-scm.org/D3860

rebase: convert opts dict to bytes at once

rebase: isolate command options from internal flags I want to get rid of per-function byteskwargs(opts).

rebase: suppress warning thrown when aborting rebase in case of dryrun Before this patch dryrun output contained "rebase aborted" every time we run rebase in dryrun mode and this warning does not sound safe from a user prespective. Differential Revision: https://phab.mercurial-scm.org/D3857

rebase: no need to store backup in case of dryrun While aborting an unfinished rebase in case of dryrun there is no need to store backup for those rebased csets. Differential Revision: https://phab.mercurial-scm.org/D3827

rebase: split _origrebase() for conveniece in dryrun This patch split _origrebase() method by extracting rbsrt part from that to make it easy to implement dryrun in more elegant way than before. Differential Revision: https://phab.mercurial-scm.org/D3856

rebase: extract dryrun as a function To avoid more number of indented blocks and make it easier to add additional functionality in dryrun, extracted as a function. Differential Revision: https://phab.mercurial-scm.org/D3855

rebase: add lock to cover whole dryrun process Before this patch it is easy for another hg to interrupt the dryrun. This patch make sure that dryrun will complete without any interruption. Differential Revision: https://phab.mercurial-scm.org/D3854

merge: add a 'keepconflictparent' argument to graft Before this change, `merge.graft` was always dropping the "grafted" changeset from the parent. This is impractical in case of conflict as the second parent can be useful to help with conflict resolution. We add a new boolean parameter to control this behavior. This will make using `merge.graft` directly in shelve practicable. Differential Revision: https://phab.mercurial-scm.org/D3692

unlinkpath: make empty directory removal optional (issue5901) (issue5826) There are known cases where performing operations such as rebase from a directory that is newly created can fail or at least lead to being in a directory handle that no longer exists. This is even reproducible by just doing something as simple as: cd foo; hg rm * The behavior is different if you use `hg addremove`, the directory is not removed until we attempt to go back to the node after committing it: cd foo; rm *; hg addremove; hg ci -m'bye foo'; hg co .^; hg co tip Differential Revision: https://phab.mercurial-scm.org/D3859

py3: convert opts keys to bytes using pycompat.byteskwargs() This fixes the py3 build broken due to 56b2074114b19e12abd9cd4c378d58b702232705. Differential Revision: https://phab.mercurial-scm.org/D3853

procutil: add a shim for translating shell commands to native commands

revset: move lookup of first ancestor() candidate out of the loop

revset: leverage orset() to flatten ancestor() arguments This also makes orset() accept an empty argument because nullary ancestor() call is valid. That's not the case for orset(), but should be okay.

revset: remove orphan i18n comment from ancestor()

grep: deprecates `--all` flag As the name "all" is a misnomer for an option that searches on diffs of revisions, we are moving to diff flag from all, deprecating it in the process. Differential Revision: https://phab.mercurial-scm.org/D3825

namespaces: let namespaces override singlenode() definition Some namespaces have multiple nodes per name (meaning that their namemap() returns multiple nodes). One such namespace is the "topics" namespace (from the evolve repo). We also have our own internal namespace at Google (for review units) that has multiple nodes per name. These namespaces may not want to use the default "pick highest revnum" resolution that we currently use when resolving a name to a single node. As an example, they may decide that `hg co <name>` should check out a commit that's last in some sense even if an earlier commit had just been amended and thus had a higher revnum [1]. This patch gives the namespace the option to continue to return multiple nodes and to override how the best node is picked. Allowing namespaces to override that may also be useful as an optimization (it may be cheaper for the namespace to find just that node). I have been arguing (in D3715) for using all the nodes returned from namemap() when resolving the symbol to a revset, so e.g. `hg log -r stable` would resolve to *all* nodes on stable, not just the one with the highest revnum (except that I don't actually think we should change it for the branch namespace because of BC). Most people seem opposed to that. If we decide not to do it, I think we can deprecate the namemap() function in favor of the new singlenode() (I find it weird to have namespaces, like the branch namespace, where namemap() isn't nodemap()'s inverse). I therefore think this patch makes sense regardless of what we decide on that issue. [1] Actually, even the branch namespace would have wanted to override singlenode() if it had supported multiple nodes. That's because closes branch heads are mostly ignored, so "hg co default" will not check out the highest-revnum node if that's a closed head. Differential Revision: https://phab.mercurial-scm.org/D3852

rebase: refactor dryrun implementation This patch refactor dry-run code to make it easy to add additional functionality in dryrun. Otherwise we had to add every functionality through _origrebase() which does not seem a good implementation. Differential Revision: https://phab.mercurial-scm.org/D3849

hooks: allow Unix style environment variables on external Windows hooks This will help making common hooks between Windows and non-Windows platforms. Having to build the shellenviron dict here and in procutil.system() is a bit unfortunate, but the only other option is to fix up the command inside procutil.system(). It seems more important that the note about the hook being run reflects what is actually run. The patch from last summer added the hooks on the command line, but it looks like HG_ARGS has since learned about --config args, and the output was just confusing. Therefore, it's now loaded from a file in the histedit test for clarity.

windows: add a method to convert Unix style command lines to Windows style This started as a copy/paste of `os.path.expandvars()`, but limited to a given dictionary of variables, converting `foo = foo + bar` to `foo += bar`, and adding 'b' string prefixes. Then code was added to make sure that a value being substituted in wouldn't itself be expanded by cmd.exe. But that left inconsistent results between `$var1` and `%var1%` when its value was '%foo%'- since neither were touched, `$var1` wouldn't expand but `%var1%` would. So instead, this just converts the Unix style to Windows style (if the variable exists, because Windows will leave `%missing%` as-is), and lets cmd.exe do its thing. I then dropped the %% -> % conversion (because Windows doesn't do this), and added the ability to escape the '$' with '\'. The escape character is dropped, for consistency with shell handling. After everything seemed stable and working, running the whole test suite flagged a problem near the end of test-bookmarks.t:1069. The problem is cmd.exe won't pass empty variables to its child, so defined but empty variables are now skipped. I can't think of anything better, and it seems like a pre-existing violation of the documentation, which calls out that HG_OLDNODE is empty on bookmark creation. Future additions could potentially be replacing strong quotes with double quotes (cmd.exe doesn't know what to do with the former), escaping a double quote, and some tilde expansion via os.path.expanduser(). I've got some doubts about replacing the strong quotes in case sh.exe is run, but it seems like the right thing to do the vast majority of the time. The original form of this was discussed about a year ago[1]. [1] https://www.mercurial-scm.org/pipermail/mercurial-devel/2017-July/100735.html

hgweb: add archive entries to graph page Changelog page has them, so it makes sense to add them to graph page too.

hgweb: add z-index for search field tooltip On graph page, search field tooltip sometimes goes down to the graph area, where it used to be covered by foreground element of graph entries (.fg) because they have z-index: 10. To prevent the tooltip from being covered, z-index: 15 is enough.

tests: pass "rev" argument to commands.update() as string commands.update() normally gets its "rev" argument as a string, but test-basic.t was passing an integer. That happened to work, but we shouldn't rely on it. Differential Revision: https://phab.mercurial-scm.org/D3851

revset: fix heads() order to always follow the input set (BC) An argument expression should never affect the order of the result set. That's the rule of the revset predicates.

test-revset: show that order of heads() can be wrong

stringutil: update list of re-special characters to include &~ I missed this because I was looking at the change that refactored re.escape, and these characters were added in https://github.com/python/cpython/commit/05cb728d68a278d11466f9a6c8258d914135c96c. Differential Revision: https://phab.mercurial-scm.org/D3850

tests: fix up some lax escaping in test-template-basic.t These misfired escapes turn into hard errors in Python 3.7, and I'd really rather we not work around it. We should *probably* try and find a way to proactively warn users about invalid escape sequences. There's one more failure of this type in this file on Python 3.7, but I can't figure out the issue. It'll need to be corrected in a follow-up. Differential Revision: https://phab.mercurial-scm.org/D3843