tests: add more testing around server.uncompressed We already have testing for server.uncompressed in test-http*.t. However, it doesn't cover the new bundle2 use case. And, we don't have comprehensive testing of advertised capabilities. We add tests to test-clone-uncompressed.t that demonstrate behavior for both legacy and bundle2 configurations. If you look closely, the bundle2 capabilities are advertising stream support when it isn't enabled. That's a bug. In addition, while the client is smart enough to not request a stream clone when the server doesn't have the feature enabled, the getbundle wire protocol command is still sending stream clone data. This doesn't match the behavior of the legacy stream_out wire protocol command. That's also a bug. Tests have been added. While I was here, I also changed how the PID is recorded in $DAEMON_PIDS. If we kill a process, the PID formerly in $DAEMON_PIDS no longer exists. So we should replace that file instead of appending to it. Differential Revision: https://phab.mercurial-scm.org/D1928

bundle2: move version of stream clone into part name I don't like having version numbers as part parameters. It means that parts can theoretically vary wildly in their generation and processing semantics. I think that a named part should have consistent behavior over time. In other words, if you need to introduce new functionality or behavior, that should be expressed by inventing a new bundle2 part, not adding functionality to an existing part. This commit applies this advice to the just-introduced stream clone via bundle2 feature. The "version" part parameter is removed. The name of the bundle2 part is now "stream2" instead of "stream" with "version=v2". Differential Revision: https://phab.mercurial-scm.org/D1927

exchange: send bundle2 stream clones uncompressed Stream clones don't compress well. And compression undermines a point of stream clones which is to trade significant CPU reductions by increasing size. Building upon our introduction of metadata to communicate bundle information back to callers of exchange.getbundlechunks(), we add an attribute to the bundler that communicates whether the bundle is best left uncompressed. We return this attribute as part of the bundle metadata. And the wire protocol honors it when determining whether to compress the wire protocol response. The added test demonstrates that the raw result from the wire protocol is not compressed. It also demonstrates that the server will serve stream responses when the feature isn't enabled. We'll address that in another commit. The effect of this change is that server-side CPU usage for bundle2 stream clones is significantly reduced by removing zstd compression. For the mozilla-unified repository: before: 37.69 user 8.01 system after: 27.38 user 7.34 system Assuming things are CPU bound, that ~10s reduction would translate to faster clones on the client. zstd can decompress at >1 GB/s. So the overhead from decompression on the client is small in the grand scheme of things. But if zlib compression were being used, the overhead would be much greater. Differential Revision: https://phab.mercurial-scm.org/D1926

tests: update test to work with Git 2.16 It looks like Git 2.16 removed the "..." from some strings. Glob over those characters in the test output. Differential Revision: https://phab.mercurial-scm.org/D1935

exchange: return bundle info from getbundlechunks() (API) We generally want a mechanism to pass information about the generated bundle back to callers (in addition to the byte stream). Ideally we would return a bundler from this function and have the caller code to an interface. But the bundling APIs are not great and getbundlechunks() is the best API we have for obtaining bundle contents in a unified manner. We change getbundlechunks() to return a dict that we can use to communicate metadata. We populate that dict with the bundle version number to demonstrate some value. .. api:: exchange.getbundlechunks() now returns a 2-tuple instead of just an iterator. Differential Revision: https://phab.mercurial-scm.org/D1925

exchange: make stream bundle part deterministic repo.requirements is a set. We need to sort it so the part content is deterministic. Differential Revision: https://phab.mercurial-scm.org/D1924

bundle2: specify what capabilities will be used for We currently assume there is a symmetric relationship of bundle2 capabilities between client and server. However, this may not always be the case. We need a bundle2 capability to advertise bundle2 streaming clone support on servers to differentiate it from the existing, legacy streaming clone support. However, servers may wish to disable streaming clone support. If bundle2 capabilities were the same between client and server, a client (which may also be a server) that has disabled streaming clone support would not be able to perform a streaming clone itself! This commit introduces a "role" argument to bundle2.getrepocaps() that explicitly defines the role being performed. This will allow us (and extensions) to alter bundle2 capabilities depending on the operation being performed. Differential Revision: https://phab.mercurial-scm.org/D1923

wireproto: don't compress errors from getbundle() Errors should be small. There's no real need to compress them. Truth be told, there's no good reason to not compress them either. But leaving them uncompressed makes it easier to test failures by looking at the raw HTTP response. This makes it easier for us to write tests. It may make it easier for people writing their own clients. Differential Revision: https://phab.mercurial-scm.org/D1922

tests: teach get-with-headers.py some new tricks We add the ability to specify arbitrary HTTP request headers and to save the HTTP response body to a file. These will be used in upcoming commits. Differential Revision: https://phab.mercurial-scm.org/D1921

tests: use argparse in get-with-headers.py I'm about to add another flag and I don't want to deal with this organic, artisanal argument parser. Differential Revision: https://phab.mercurial-scm.org/D1920

convert: use a collections.deque This function was doing a list.pop(0) on a list whose length was the number of revisions to convert. Popping an early element from long lists is not an efficient operation. collections.deque supports efficient inserts and pops at both ends. So we switch to that data structure. When converting the mozilla-unified repository, which has 445,748 revisions, this change makes the "sorting..." step of `hg convert --sourcesort` significantly faster: before: ~59.2s after: ~1.3s Differential Revision: https://phab.mercurial-scm.org/D1934

repair: invalidate volatile sets after stripping Matt Harbison reported that some tests were broken on Windows after 1a09dad8b85a (evolution: report new unstable changesets, 2018-01-14). The failures were exactly as seen in this patch. The failures actually seemed correct, which made me wonder why they didn't fail the same way on Linux. It turned out to be a cache invalidation problem. The new orphan mentioned in the test case actually does get created when we're re-applying the temporary bundle that's created while stripping. However, without the invalidation, it appears that there was already an orphan before applying the temporary bundle. The warnings about unknown working parent appear because the aformentioned changeset means that we're now accessing the dirstate while it's invalid. We may want to suppress these messages that happen in the intermediate strip state, but they're technically correct (although confusing to the user), so I think just fixing the cache invalidation is fine for now. I haven't figured out why the caches seemed to get correctly invalidated on Windows. Differential Revision: https://phab.mercurial-scm.org/D1933

subrepo: handle 'C:' style paths on the command line (issue5770) If you think 'C:' and 'C:\' are equivalent paths, see the inline comment before proceeding. The problem here was that several commands that take a URL argument (incoming, outgoing, pull, and push) will use that value to set 'repo._subtoppath' on the repository object after command specific manipulation of it, but before converting it to an absolute path. When an operation is performed on a relative subrepo, subrepo._abssource() will posixpath.join() this value with the relative subrepo path. That adds a '/' after the drive letter, changing how it is evaluated by abspath()/realpath() in vfsmod.vfs(..., realpath=True) as the subrepo is instantiated. I initially tried sanitizing the path in url.localpath(), because url.isabs() only checks that it starts with a drive letter. By the sample behavior, this is clearly not an absolute path. (Though the comment in isabs() is weasely- this style path can't be joined either.) But not everything funnels through there, and it required explicitly calling localpath() in hg.parseurl() and assigning to url.path to fix. But then tests failed with urls like 'a#0'. Next up was sanitizing the path in the url constructor. That caused doctest failures, because there are drive letter tests, so those got expanded in system specific ways. Yuya correctly pointed out that util.url is a parser, and shouldn't be substituting the path too. Rather than fixing every command call site, just convert it in the common subrepo location. I don't see any sanitizing on the path config options, so I fixed those too. Note that while the behavior is fixed here, there are still places where 'comparing with C:' gets printed out, and that's not great for debugging purposes. (Specifically I saw it in `hg incoming -B C:`, without subrepos.) While clone will write out an absolute default path, I wonder what would happen if a user edited that path to be 'C:'. (I don't think supporting relative paths in .hgrc is a sane thing to do, but while we're poking holes in things...) Since this is such an oddball case, it still leaks through in places, and there seems to be a lot of duplicate url parsing, maybe the url parsing should be moved to dispatch, and provide the command with a url object? Then we could convert this to an absolute path once, and not have to worry about it in the rest of the code. I also checked '--cwd C:' on the command line, and it was previously working because os.chdir() will DTRT. Finally, one other note from the url.localpath() experimenting. I don't see any cases where 'self._hostport' can hold a drive letter. So I'm wondering if that is wrong/old code.

dummysmtpd: don't die on client connection errors The connection refused error in test-patchbomb-tls.t[1] is sporadic, but one of the more often seen errors on Windows. I added enough logging to a file and dumped it out at the end to make the following observations: - The listening socket is successfully created and bound to the port, and the "listening at..." message is always logged. - Generally, the following is the entire log output, with the "accepted ..." message having been added after `sslutil.wrapserversocket`: listening at localhost:$HGPORT $LOCALIP ssl error accepted connect accepted connect $LOCALIP from=quux to=foo, bar $LOCALIP ssl error - In the cases that fail, asyncore.loop() in the run() method is exiting, but not with an exception. - In the cases that fail, the following is logged right after "listening ...": Traceback (most recent call last): File "c:\\Python27\\lib\\asyncore.py", line 83, in read obj.handle_read_event() File "c:\\Python27\\lib\\asyncore.py", line 443, in handle_read_event self.handle_accept() File "../tests/dummysmtpd.py", line 80, in handle_accept conn = sslutil.wrapserversocket(conn, ui, certfile=self._certfile) File "..\\mercurial\\sslutil.py", line 570, in wrapserversocket return sslcontext.wrap_socket(sock, server_side=True) File "c:\\Python27\\lib\\ssl.py", line 363, in wrap_socket _context=self) File "c:\\Python27\\lib\\ssl.py", line 611, in __init__ self.do_handshake() File "c:\\Python27\\lib\\ssl.py", line 840, in do_handshake self._sslobj.do_handshake() error: [Errno 10054] $ECONNRESET$ - If the base class handler is overridden completely, the the first "ssl error" line is replaced by the stacktrace, but the other lines are unchanged. The client behaves no differently, whether or not the server stacktraced. In general, `./run-tests.py --local -j9 -t9000 test-patchbomb-tls.t --runs-per-test 20` would show the issue after a run or two. With this change, `./run-tests.py --local -j9 -t9000 test-patchbomb-tls.t --loop` ran 800 times without a hiccup. This makes me wonder if the other connection refused messages that bubble up on occasion are caused by a similar issue. It seems a bit drastic to kill the whole server on account of a single communication failure with a client. # no-check-commit because of handle_error() [1] https://buildbot.mercurial-scm.org/builders/Win7%20x86_64%20hg%20tests/builds/421/steps/run-tests.py%20%28python%202.7.13%29/logs/stdio

cext: define MIN macro only if it is not yet defined MIN macro is defined in <sys/param.h> on macOS Sierra. Therefore as HAVE_BSD_STATFS is defined in osutil.c, 'MIN' macro redefined warning is emitted.

copyright: update to 2018 January seems to be a good month to do this.

tests: glob copyright years in test-extension.t Other tests already do this.

test-sshserver: stabilize for Windows

test-branch-change: stabilize for Windows

test-sparse: make the '.hg' exclusion filter Windows compatible

lfs: rename {lfsattrs} to {pointer} This seems more descriptive.

lfs: expand the user facing documentation

streamclone: also stream caches to the client When stream clone is used over bundle2, relevant cache files are also streamed. This is expected to be a massive performance win for clone since no important cache will have to be recomputed. Some performance numbers: (All times are wall-clock times in seconds, 2 attempts per case.) # Mozilla-Central ## Clone over ssh over lan V1 streaming: 234.3 239.6 V2 streaming: 248.4 243.7 ## Clone over ssh over Internet V1 streaming: 175.5 110.9 V2 streaming: 109.1 111.0 ## Clone over HTTP over lan V1 streaming: 105.3 105.6 V2 streaming: 112.7 111.4 ## Clone over HTTP over internet V1 streaming: 105.6 114.6 V2 streaming: 226.7 225.9 ## Hg tags V1 streaming (no cache): 1.084 1.071 V2 streaming (cache): 0.312 0.325 ## Hg branches V1 streaming (no cache): 14.047 14.148 V2 streaming (with cache): 0.312 0.333 # Pypy ## Clone over ssh over internet V1 streaming: 29.4 30.1 V2 streaming: 31.2 30.1 ## Clone over http over internet V1 streaming: 29.7 29.7 V2 streaming: 75.2 72.9 (since ssh and lan are not affected, there seems to be an issue with how we read/write the http stream on connection with latency, unrelated to the format) ## Hg tags V1 streaming (no cache): 1.752 1.664 V2 streaming (with cache): 0.274 0.260 ## Hg branches V1 streaming (no cache): 4.469 4.728 V2 streaming (with cache): 0.318 0.321 # Private repository: * 500K revision revisions * 11K topological heads * 28K branch heads ## hg tags no cache: 1543.332 with cache: 4.900 ## hg branches no cache: 91.828 with cache: 2.955

caches: make 'cachetocopy' available in scmutil For more code to use this information, we need it to be more publicly available.

streamclone: add support for cloning non append-only file The phaseroots are stored in a non append-only file in the repository. We include them in the stream too. Since they are not append-only, we have to keep a copy around while we hold the lock to be able to stream them later. Since phase get exchanged within the stream we can skip requesting them independently. As a side effect, this will fixes issue5648 once the feature is enabled by default.

streamclone: tests phase exchange during stream clone We add a test dedicated to phases. As reported in issue 5648 stream from a non publishing server is currently broken (does not preserve the phase). We'll fix it with 'v2' support in the next changesets.

streamclone: add support for bundle2 based stream clone The feature put to use the various bits introduced previously. If the server supports it, the client will request its stream clone through bundle2 instead of the legacy 'stream_out' commands. The bundle2 version use the better 'v2' version of stream bundles. The 'v2' format is not finalized yet. Now that there are some code running it, we can start working on it again. Performance numbers are available at the end of this series.

pull: preindent some code Next changesets will add support for using stream cloning with bundle2. We introduce indentation change first for clarity.

pull: reorganize bundle2 argument bundling We are about to add the ability to use stream bundle with bundle2. Before doing so, we need to gather some code that will not be used in the bundle2 case. There is no behavior change within this changeset.

clone: allow bundle2's stream clone with 'server.disablefullbundle' The previous check was a bit too strict and would not recognize a get bundle not requesting changegroup.

bundle2: add support for a 'stream' parameter to 'getbundle' This parameter can be used to request a stream bundle.

bundle2: add a 'stream' part handler for stream cloning The part contains the necessary arguments and payload to handle a stream bundle v2. It will be put to use in later changesets.

streamclone: rework canperformstreamclone There is code about bundle2 laying around in `canperformstreamclone` but not put to any uses. As we discovered with the previous patch, streambundle 'v1' won't work on bundle2 because they are readline based. So we jump to 'v2' as the first expected supported version.

streamclone: define first iteration of version 2 of stream format (This patch is based on a first draft from Gregory Szorc, with deeper rework) Version 1 of the stream clone format was invented many years ago and suffers from a few deficiencies: 1) Filenames are stored in store-encoded (on filesystem) form rather than in their internal form. This makes future compatibility with new store filename encodings more difficult. 2) File entry "headers" consist of a newline of the file name followed by the string file size. Converting strings to integers is avoidable overhead. We can't store filenames with newlines (manifests have this limitation as well, so it isn't a major concern). But the big concern here is the necessity for readline(). Scanning for newlines means reading ahead and that means extra buffer allocations and slicing (in Python) and this makes performance suffer. 3) Filenames aren't compressed optimally. Filenames should be compressed well since there is a lot of repeated data. However, since they are scattered all over the stream (with revlog data in between), they typically fall outside the window size of the compressor and don't compress. 4) It can only exchange stored based content, being able to exchange caches too would be nice. 5) It is limited to a stream-based protocol and isn't suitable for an on-disk format for general repository reading because the offset of individual file entries requires scanning the entire file to find file records. As part of enabling streaming clones to work in bundle2, #2 proved to have a significant negative impact on performance. Since bundle2 provides the opportunity to start fresh, Gregory Szorc figured he would take the opportunity to invent a new streaming clone data format. The new format devised in this series addresses #1, #2, and #4. It punts on #3 because it was complex without yielding a significant gain and on #5 because devising a new store format that "packs" multiple revlogs into a single "packed revlog" is massive scope bloat. However, this v2 format might be suitable for streaming into a "packed revlog" with minimal processing. If it works, great. If not, we can always invent stream format when it is needed. This patch only introduces the bases of the format. We'll get it usable through bundle2 first, then we'll extend the format in future patches to bring it to its full potential (especially #4).

util: implement varint functions This will be useful in an incoming version-2 of the stream format.

util: move 'readexactly' in the util module This function is used in multiple place, having it in util would be better. (existing caller will be migrated in another series)

lfs: separate a debug message from the subsequent abort message

sshserver: add a couple of tests for argument parsing I noticed that we didn't have any unit tests covering wire protocol argument parsing.

merge with stable

wireproto: split streamres into legacy and modern case A couple of commands currently require transmission of uncompressed frames with the old MIME type. Split this case from streamres into a new streamres_legacy class. Streamline the remaining code accordingly. Add a new flag to streamres to request uncompressed streams. This is useful for sending data that is already compressed like a pre-built bundle. Expect clients to support uncompressed data. For older clients, zlib will still be used. Differential Revision: https://phab.mercurial-scm.org/D1862

localrepo: run cache-warming transaction callback before report callback See in-code comment for details. Differential Revision: https://phab.mercurial-scm.org/D1918

scmutil: 0-pad transaction report callback category Before this patch, the transaction name was '%2i-txnreport', which means the first one would be ' 0-txnreport'. It seems more intuitive for sorting purposes (the callbacks are called in lexicographical order) to make it 0-padded. Differential Revision: https://phab.mercurial-scm.org/D1917

stream: add a test showing we also clone bookmarks Bookmarks are not stored in `.hg/store`. We need to make sure they are cloned with `--stream`.

branch: allow changing branch name to existing name if possible With the functionality added in previous patch we can change branches of a revision but not everytime even if it's possible to do so. For example cosider the following case: o 3 added a (foo) o 2 added b (foo) o 1 added c (bar) o 0 added d (bar) Here if I want to change the branch of rev 2,3 to bar, it was not possible and it will say, "a branch with same name exists". This patch allows us to change branch of 2,3 to bar. The underlying logic for changing branch finds the changesets from the revs passed which have no parents in revs. We only support revsets which have only one such root, so to support this we check whether the parent of the root has the same name as that of the new name and if so, we can use the new name to change branches. Differential Revision: https://phab.mercurial-scm.org/D1913

branch: add a --rev flag to change branch name of given revisions This patch adds a new --rev flag to hg branch which can be used to change branch of revisions. This is motivated from topic extension where you can change topic on revisions but this one has few restrictions which are: 1) You cannot change branch name in between the stack 2) You cannot change branch name and set it to an existing name 3) You cannot change branch of non-linear set of commits 4) You cannot change branch of merge commits. Tests are added for the same. .. feature:: An experimental flag `--rev` to `hg branch` which can be used to change branch of changesets. Differential Revision: https://phab.mercurial-scm.org/D1074

templater: fix crash by empty group expression 'exp' may be None because of '(group None)' node. The error message is copied from revset.py.

log: fix typo in comment about _matchfiles()

sparse: --include 'dir1/dir2' should not include 'dir1/*' In 2015 there was a workaround added (f39bace2d6cad32907c0d7961b3c0dbd64a1b7ad) to sparse in the hg-experimental repo. That workaround: a) no longer seems to be needed, since its testcase passes even with the code removed, and b) caused a new problem: --include 'dir1/dir2' ended up including dir1/* too. (--include 'glob:dir1/dir2' is a user-level workaround.) Remove the offending code, and add a testcase for situation B.

fileset: add kind:pat operator ":" isn't taken as a symbol character but an infix operator so we can write e.g. "path:'foo bar'" as well as "'path:foo bar'". An invalid pattern kind is rejected in the former form as we know a kind is specified explicitly. The binding strength is copied from "x:y" range operator of revset. Perhaps it can be adjusted later if we want to parse "foo:bar()" as "(foo:bar)()", not "foo:(bar())". We can also add "kind:" postfix operator if we want. One possible confusion is that the scope of the leading "set:" vs "kind:pat" operator. The former is consumed by a matcher so applies to the whole fileset expression: $ hg files 'set:foo() or kind:bar or baz' ^^^^^^^^^^^^^^^^^^^^^^^^ Whereas the scope of kind:pat operator is narrow: $ hg files 'set:foo() or kind:bar or baz' ^^^

minifileset: unify handling of symbol and string patterns They must be identical for fileset compatibility.

fileset: move import of match module to top Actually there was no circular import issue.

revlog: group delta computation methods under _deltacomputer object Extracting these methods from revlog will allow changing the implementation of the deltacomputer, by providing this interface: __init__(self, revlog) - constructor that initialize the object from a given revlog buildtext(self, revinfo, fh) - builds the fulltext version of a revision from a _revisioninfo object and the file handle to the .d (or .i for inline mode) file. finddeltainfo(self, revinfo, fh) - find a revision in the revlog against which it is acceptable to build a delta, and build the corresponding _deltainfo. It should now be easier to write an experimental feature that would replace _deltacomputer by another object, for example one that would know how to parallelize the delta computation in order to quicken the storage of multiple revisions.

revlog: refactor out _finddeltainfo from _addrevision Splicing the code into smaller chunks should help understanding it, and eventually override some parts in experimental branches to try optimization.

localrepo: micro-optimize __len__() to bypass repoview Since len(changelog) isn't overridden, we don't have to validate a cache of unfiltered changelog. $ python -m timeit -n 10000 \ -s 'from mercurial import hg, ui; repo = hg.repository(ui.ui());' \ 'len(repo)' orig) 10000 loops, best of 3: 32.1 usec per loop new) 10000 loops, best of 3: 1.79 usec per loop Spotted by Jordi Gutiérrez Hermoso.

lfs: defer registering the pre-push hook until blobs are committed The hook searches outgoing commits for blobs, and uploads them before letting the push occur. No reason to search for that which isn't there.

lfs: dump the full response on httperror in debug mode This was immensely helpful in diagnosing the 500: Internal Server Error when using workers to upload. It's a nasty wall of html, so we really can't do anything else with it.

lfs: default the User-Agent header for blob transfers to 'git-lfs' The custom User-Agent for blob transfers was added in e7bb5fc4570c. Now I've hit another incompatibility with a server wanting the string to start with 'git' or 'git-lfs' [1]. I don't feel strongly about this either way, but a Wireshark trace of git shows that when the Batch API is hit, the User-Agent is 'git-lfs/2.3.4'. So this would probably ensure maximum interoperability. This still leaves the experimental knob in, just in case. [1] https://bitbucket.org/sdorra/scm-manager/src/095a027178888bc2b819aebfae3d2c192c858030/scm-plugins/scm-git-plugin/src/main/java/sonia/scm/web/GitUserAgentProvider.java?at=default&fileviewer=file-view-default#GitUserAgentProvider.java-117

lfs: default to not using workers for upload/download I ran into truncated uploads with this defaulting to on. Wojciech Lis diagnosed it as creating keepalive connections prior to forking, and illegally multiplexing the same connection. [1] I didn't notice a problem with the couple of downloads I tried, but disabled both for simplicity and safety. [1] https://www.mercurial-scm.org/pipermail/mercurial-devel/2018-January/109916.html

lfs: add the '{lfsattrs}' template keyword to '{lfs_files}' This provides access to the metadata dictionary contained within the tracked pointer file. The OID is probably the most important attribute, and has its own keyword. But we might as well have this for completeness. I liked {pointer} better, but couldn't make it work with the singular/plural forms.

debugdownload: read repository hgrc if there is one The command does not require a repository, but will use it if there is one. This simplifies the reading of the remote destination when testing for largefile based url.

test-blackbox: stabilize for Windows This goes with 853bf7d90804.

debugdeltachain: cleanup the double call to _slicechunk Follow-up to Yuya's review on 43154a76f3927c4f0c8c6b02be80f0069c7d8fdb: > Nit: hasattr() isn't necessary. revlog._slicechunk() is used in the previous > block. hasattr() isn't necessary indeed, as we are protected by the withsparseread option, which was introduced at the same time as revlog._slicechunk, in e2ad93bcc084b97c48f54c179365376edb702858. And, as Yuya noticed, _slicechunk could be called only once.

blackbox: don't unpack the list while passing into str.join() The current state may result in error TypeError. Caught using evolve-tests.

atomicupdate: add an experimental option to use atomictemp when updating In some cases Mercurial truncating files when updating causes problems. It can happens when processes are currently reading the file or with big file or on NFS mounts. We add an experimental option to use the atomictemp option of vfs.__call__ in order to avoid the problem. The localrepository.wwrite seems to assume the files are created without the `x` flag; with atomictempfile, the new file might inherit the `x` flag from the destination. We force remove it afterward. This code could be refactored and the flag processing could be moved inside vfs. This patch should be tested with `--extra-config-opt experimental.update.atomic-file=True` as we disabled the option by default. Differential Revision: https://phab.mercurial-scm.org/D1882