transaction: split new files into a separate set Journal entries with size 0 are common as they represent new revlog files. Move them from the dictionary into a set as the former is more dense. This reduces peak RSS by 70MB for the NetBSD test repository with around 450k files under .hg/store. Differential Revision: https://phab.mercurial-scm.org/D9278

transaction: change list of journal entries into a dictionary The transaction object used to keep a mapping table of path names to journal entries and a list of journal entries consisting of path and file offset to truncate on rollback. The offsets are used in three cases. repair.strip and rollback process all of them in one go, but they care about the order. For them, it is perfectly reasonable to read the journal back from disk as both operations already involve at least one system call per journal entry. The other consumer is the revlog logic for moving from inline to external data storage. It doesn't care about the order of the journal and just needs to original offset stored. Further optimisations are possible here to move the in-memory journal to a set(), but without memoisation of the original revlog size this could turn it into O(n^2) behavior in worst case when many revlogs need to migrated. Differential Revision: https://phab.mercurial-scm.org/D9277

transaction: rename find to findoffset and drop backup file support transaction.find used to support access to both the regular file and backup file list. They have different formats, so any consumer has to be aware of the difference alredy. There is no in-core consumer for the backup file access, so don't provide it. Differential Revision: https://phab.mercurial-scm.org/D9276

transaction: drop per-file extra data support At the moment, transactions support an optional extra data argument for all files to be stored in addition to the original offset. This is used in core only by the revlog inline to external data migration. It is used to memoize the number of revisions before the transaction. That number of can be computed during the walk easily, so drop the requirement. Differential Revision: https://phab.mercurial-scm.org/D9275

templates: define a {onelinesummary} keyword It is sometimes useful to be able to use the configured `command-template.oneline-summary` in higher-level templates. For example, I would like to use it in an internal template that lists commits in a "review unit" (kind of a pull request). This patch adds support for that. We may want to define a way of formatting a context using a command-specific override (from `command-templates.oneline-summary.<command>`), but that will have to be a template function instead. I don't plan to do that, but I'm mentioning it now in case reviewers would prefer that we use a no-arg function (i.e. `{onelinesummary()}`) already today to prepare for that. Differential Revision: https://phab.mercurial-scm.org/D9314

relnotes: document new [command-templates] section Differential Revision: https://phab.mercurial-scm.org/D9266

help: document the new [command-templates] config section Differential Revision: https://phab.mercurial-scm.org/D9265

strip: move into core As discussed at the 5.2 sprint, replace strip extension by a core command, debugstrip. Obviously, the extension stays for backwards compatibility. As an implementation note, I moved the strip file as is into core, which is not done elsewhere, AFAIK. I could have inlined it into debugcommands, but that doesn't sound great. Differential Revision: https://phab.mercurial-scm.org/D9285

revlog: pass sidedata argument to flagutil.processflagswrite() Bug found through pytype. Differential Revision: https://phab.mercurial-scm.org/D9280

pure: guard against empty blocks If blocks is empty, we append `None` to the returned list, which is incorrect. This subtle issue was caught by pytype, which correctly identified the return value as List[Optional[Tuple]] because of this possibility. Differential Revision: https://phab.mercurial-scm.org/D9279

rust-status: don't bubble up os errors, translate them to bad matches In the rare cases when either the OS/filesystem throws an error on an otherwise valid action, or because a path is not representable on the filesystem, or because of concurrent actions in the filesystem, we want to warn the user about said path instead of bubbling up the error, causing an exception to be raised in the Python layer. Differential Revision: https://phab.mercurial-scm.org/D9320

rust-status: properly translate OSError to Python This is probably never going to be called after the next few patches, but we might as well make sure this is done correctly for the future rewrite. Differential Revision: https://phab.mercurial-scm.org/D9319

shelve: clear merge state after partial shelve Differential Revision: https://phab.mercurial-scm.org/D9335

tests: show that interactive shelve can leave the repo with a merge state If part of a file is shelved (as we already do in a test), there will be an unfinished merge state left after `hg shelve` finishes. There should never be a merge conflict and there should never be a reason that the user would like to re-resolve conflicts, so we should clear that state (see next patch). Differential Revision: https://phab.mercurial-scm.org/D9334

histedit: disable color while rendering template for use in plan Differential Revision: https://phab.mercurial-scm.org/D9324

tests: show how `hg histedit` can put color codes in histedit plan Differential Revision: https://phab.mercurial-scm.org/D9323

split: disable color while rendering template for use in commit message Differential Revision: https://phab.mercurial-scm.org/D9322

tests: show how `hg split` can put color codes in commit template With D9255, I made it so `hg split` respects the `commmand-templates.oneline-summary` config. I don't think I realized that the output I modified was being put in a commit message template. The result was that if you have coloring enabled, you get colors in the commit template. This patch show that. The test is unfortunately pretty verbose (like most other `hg split` tests) and shows a bunch of irrelevant "color codes" (templater labels). Differential Revision: https://phab.mercurial-scm.org/D9321

dispatch: move some helper functions down into scmutil I plan to reuse `formatparse()` in the next patch. Differential Revision: https://phab.mercurial-scm.org/D9331

errors: raise more specific errors from rewriteutil Differential Revision: https://phab.mercurial-scm.org/D9330

chgserver: backport py3 buffered I/O workarounds from procutil I've recently switched to new machine and I found chg's stdout is fully buffered. Even though chg server is a daemon process, it inherits the environment where the chg client originally forked the server. This means the server's stdout might have been wrapped by LineBufferedWrapper. That's why we need to do wrap/unwrap in both ways. The "if" condition in _restoreio() looks weird, but I'm not willing to clean things up because stdio behavior is fundamentally different between py2 and py3, and py2 support will be dropped anyway.

chg: reset errno prior to calling strtol() Otherwise we can't figure out if the last strtol() invocation failed or not.

chg: do not close dir fd while iterating It works so long as the dp is the last entry, but readdir(dp) would fail with EBADF. Let's not do that and close the dir fd explicitly.

chg: show debug message for each fd to be closed It helps debugging. The number of file descriptors should be small in most cases, so the console output wouldn't get bloated even with CHG_DEBUG=1.

chg: apply clang-format

errors: use InputError for some errors on `hg clone` Differential Revision: https://phab.mercurial-scm.org/D9329

errors: raise InputError when given non-existent paths etc Differential Revision: https://phab.mercurial-scm.org/D9328

errors: use InputError for errors about bad label names (tags etc) Differential Revision: https://phab.mercurial-scm.org/D9327

errors: use InputError for errors about bad paths Differential Revision: https://phab.mercurial-scm.org/D9326

destutil: raise more specific error when histedit.defaultrev is empty Differential Revision: https://phab.mercurial-scm.org/D9313

errors: raise more specific errors when default remote not configured Differential Revision: https://phab.mercurial-scm.org/D9312

errors: set detailed exit code to 30 for config errors This is per https://www.mercurial-scm.org/wiki/ErrorCategoriesPlan. Differential Revision: https://phab.mercurial-scm.org/D9311

errors: introduce StateError and use it from commands and cmdutil This very similar to an earlier patch (which was for `InputError`). In this patch, I also updated the transplant extension only because `test-transplant.t` would otherwise have needed a `#if continueflag`. Differential Revision: https://phab.mercurial-scm.org/D9310

errors: set detailed exit code to 100 for some remote errors This is per https://www.mercurial-scm.org/wiki/ErrorCategoriesPlan. Differential Revision: https://phab.mercurial-scm.org/D9309

errors: catch urllib errors specifically instead of using safehasattr() Before this patch, we would catch `IOError` and `OSError` and check if the instance had a `.code` member (indicates `HTTPError`) or a `.reason` member (indicates the more generic `URLError`). It seems to me that can simply catch those exception specifically instead, so that's what this code does. The existing code is from fbe8834923c5 (commands: report http exceptions nicely, 2005-06-17), so I suspect it's just that there was no `urllib2` (where `URLError` lives) back then. The old code mentioned `SSLError` in a comment. The new code does *not* try to catch that. The documentation for `ssl.SSLError` says that it has a `.reason` property, but `python -c 'import ssl; print(dir(ssl.SSLError("foo", Exception("bar"))))` doesn't mention that property on either Python 2 or Python 3 on my system. It also seems that `sslutil` is pretty careful about converting `ssl.SSLError` to `error.Abort`. It also is carefult to not assume that instances of the exception have a `.reason`. So I at least don't want to catch `ssl.SSLError` and handle it the same way as `URLError` because that would likely result in a crash. I also wonder if we don't need to handle it at all (because `sslutil` might handle all the cases). It's now early in the release cycle, so perhaps we can just see how it goes? Differential Revision: https://phab.mercurial-scm.org/D9318

errors: raise InputError in fancyopts If a value of wrong type is passed to a command line flag, that's cleary an InputError. Differential Revision: https://phab.mercurial-scm.org/D9308

packaging: switch centos 7 packaging to python 3 Differential Revision: https://phab.mercurial-scm.org/D9293

packaging: remove centos5 and centos6 support Differential Revision: https://phab.mercurial-scm.org/D9292

test-filecache: use sys.executable to call python As was mentioned in c102b704edb5, test scripts calling 'python' or 'python3' might use the wrong python. For test-filecache.py, this causes a failed test on CentOS 7. Differential Revision: https://phab.mercurial-scm.org/D9295

make: add a pyoxidizer target Differential Revision: https://phab.mercurial-scm.org/D9291

pyoxidizer: switch to modern config using run_command instead of run_mode Differential Revision: https://phab.mercurial-scm.org/D9290

pyoxidizer: default to one-file binary on non-Windows platforms Windows has some extra constraints that require a multi-file install, but we expect folks to use an MSI or similar installer there so it's less of a big deal. Differential Revision: https://phab.mercurial-scm.org/D9289

global: use python3 in shebangs Python 3 is the future. We want Python scripts to be using Python 3 by default. This change updates all `#!/usr/bin/env python` shebangs to use `python3`. Does this mean all scripts use or require Python 3: no. In the test environment, the `PATH` environment variable in tests is updated to guarantee that the Python executable used to run run-tests.py is used. Since test scripts all now use `#!/usr/bin/env python3`, we had to update this code to install a `python3` symlink instead of `python`. It is possible there are some random scripts now executed with the incorrect Python interpreter in some contexts. However, I would argue that this was a pre-existing bug: we should almost always be executing new Python processes using the `sys.executable` from the originating Python script, as `python` or `python3` won't guarantee we'll use the same interpreter. Differential Revision: https://phab.mercurial-scm.org/D9273

tests: use python from environment in test-parseindex2.py Without this, the test starts failing with D9273 (the change to `pyexename` to be specific). Differential Revision: https://phab.mercurial-scm.org/D9286

errors: set detailed exit code to 20 for locking errors This is per https://www.mercurial-scm.org/wiki/ErrorCategoriesPlan. Differential Revision: https://phab.mercurial-scm.org/D9242

errors: introduce InputError and use it from commands and cmdutil This patch introduces a `InputError` class and replaces many uses of `error.Abort` by it in `commands` and `cmdutil`. This is a part of https://www.mercurial-scm.org/wiki/ErrorCategoriesPlan. There will later be a different class for state errors (to raise e.g. when there's an unfinished operation). It's not always clear when one should report an input error and when it should be a state error. We can always adjust later if I got something wrong in this patch (but feel free to point out any you notice now). Differential Revision: https://phab.mercurial-scm.org/D9167

errors: add config that lets user get more detailed exit codes This adds an experimental config that lets the user get more detailed exit codes. For example, there will be a specific error code for input/user errors. This is part of https://www.mercurial-scm.org/wiki/ErrorCategoriesPlan. I've made the config part of tweakdefaults. I've made the config enabled by default in tests. My reasoning is that we want to see that each specific error case gives the right exit code and we don't want to duplicate all error cases in the entire test suite. It also makes it easy to grep the `.t` files for `[255]` to find which cases we have left to fix. The logic for the current exit codes is quite simple, so I'm not too worried about regressions there. I've added a test case specifically for the "legacy" exit codes. I've set the detailed exit status only for the case of `InterventionRequired` and `SystemExit` for now (the cases where we currently return something other than 255), just to show that it works. Differential Revision: https://phab.mercurial-scm.org/D9238

worker: raise exception instead of calling sys.exit() with child's code When a worker process returns an error code, we would call `sys.exit()` with that exit code on the main process. The `SystemExit` exception would then get caught in `scmutil.callcatch()`, which would return that error code. The comment there says "Commands shouldn't sys.exit directly", which I agree with. This patch changes it so we raise a specific exception when a worker fails so we can catch instead. I think that means that `SystemExit` is now always an internal error. (I had earlier thought that this call to `sys.exit()` was from within the child process until Matt Harbison made me look again, so thanks for that!) Differential Revision: https://phab.mercurial-scm.org/D9287

config: read system hgrc in lexicographical order This is similar to edbcf5b239f9 (config: read configs from directories in lexicographical order, 2019-04-03). Apparently I forgot to sort the system hgrc files there. That's fixed by this patch. Differential Revision: https://phab.mercurial-scm.org/D9269

relnotes: drop 5.6 release entries from next Differential Revision: https://phab.mercurial-scm.org/D9282

merge with stable

makefile: use Python 3 by default (BC) This change is long overdue IMO. .. bc:: Makefile now uses `python3` instead of `python` by default on non-Windows platforms. This means Mercurial will be built and run with Python 3 instead of Python 2.7 by default. To continue using Python 2, set the PYTHON variable. e.g. `make install PYTHON=python2.7`. Differential Revision: https://phab.mercurial-scm.org/D7258

hgweb: don't call sys.exit() in httpservice.run() If I'm reading the code correctly, `mercurial.server.createservice()` can return an hgweb service or one of three types of command server services. The caller then calls `mercurial.server.runservice()`, passing it the returned service's run method. Only the hgweb service was calling `sys.exit()`. It has been that way since 8d44649df03b (refactor ssh server., 2006-06-04). That commit message doesn't provide any explanation. Let's clean up and have the code follow the usual return path into the `dispatch` module. After this patch, there should be no remaining places left where we call `sys.exit()` except for valid uses in the `dispatch` and `worker` modules. Differential Revision: https://phab.mercurial-scm.org/D9272

serve: simply return instead of calling sys.exit() in `hg serve --stdio` The shouldn't be a reason to call `sys.exit()` instead of letting the code return normally. I've remove the call in both `hg serve` and `hg debugserve`. Differential Revision: https://phab.mercurial-scm.org/D9271

httpservice: move sys.exit() out of serve_forever() This is a simple refactoring to show the callers of the method, so it's easier to reason about the impact of removing the `sys.exit()` calls in subsequent patches. Differential Revision: https://phab.mercurial-scm.org/D9270

copies: handle more cases where a file got replaced by a copy This patch fixes the changeset-centric version in a pretty straight-forward way. It fixes it to automatically resolve the conflict, which is better than resulting in a modify/delete conflict as it was before b4057d001760 (merge: when rename was made on both sides, use ancestor as merge base, 2020-01-22). I'll leave it for later to test and explicitly handle cases where files have been renamed to the same target on different sides of the merge. Differential Revision: https://phab.mercurial-scm.org/D8653

tests: test more cases where a file got replaced by a copy This adds a test where a file is modified on one branch and is renamed onto another file in another branch. That should ideally be automatically resolved (by propagating the modification to the rename destination). Alternatively, it could be considered a modify/delete conflict. It should at least not be automatically resolved by ignoring the modification. However, that is what actually happens with the changeset-centric algorithm since I broke it in b4057d001760 (merge: when rename was made on both sides, use ancestor as merge base, 2020-01-22). Before that commit, it resulted in a modify/delete conflict. The filelog-centric algorithm was broken already before that commit. Differential Revision: https://phab.mercurial-scm.org/D8652

unionrepo: don't insert index tuples with None as int field None is not a valid size. Use -1 as placeholder instead. This will be necessary when the index starts enforcing type correctness. Differential Revision: https://phab.mercurial-scm.org/D9161

bundlerepo: don't insert index tuples with full nodes as linkrev The index format has a documented format and latter changes will start to enforce the field types. The bundlerepo uses full nodes for the linkrev field when it should be using revision numbers. Use the link mapping to resolve them, except in the special case of self-references. Those are actually indications of a missing linkrev. Differential Revision: https://phab.mercurial-scm.org/D9160

rhg: add full node id support for `debugdata` command Unlike other later implemented commands `debugdata` only supported revision number. This changeset add full node id support for consistency with other commands. Differential Revision: https://phab.mercurial-scm.org/D9230