merge: remove direct rustmod reference We shouldn't rely on this member being present in `dirstate.py`, this creates unnecessary coupling. This also can trigger certain issues in edge-cases where the policy is changed at runtime or multiple Python environments fight, which is an added bonus. Differential Revision: https://phab.mercurial-scm.org/D12217

relnotes: add 6.1 Differential Revision: https://phab.mercurial-scm.org/D12205

relnotes: add 6.0.3 Differential Revision: https://phab.mercurial-scm.org/D12204

windows: adjust test output in test-merge-tools.t Differential Revision: https://phab.mercurial-scm.org/D12216

windows: generalize output for test-status-tracked-key.t Differential Revision: https://phab.mercurial-scm.org/D12215

windows: skip a section of a test that is legitimately broken on windows See the comment in the test itself. Differential Revision: https://phab.mercurial-scm.org/D12214

heptapod-ci: remove Python 2 support (hurray!) 6.1 was the last release to support Python 2, let's lighten the CI. Differential Revision: https://phab.mercurial-scm.org/D12206

Added signature for changeset c00d3ce4e94b

Added tag 6.1rc0 for changeset c00d3ce4e94b

branching: merge default into stable for 6.1 freeze

branching: merge 6.0.3 stable into default

branching: merge stable into default

tracked-key: remove the dual write and rename to tracked-hint The dual-write approach was mostly useless. As explained in the previous version of the help, the key had to be read twice before we could cache a value. However this "read twice" limitation actually also apply to any usage of the key. If some operation wants to rely of the "same value == same tracked set" property it would need to read the value before, and after running that operation (or at least, after, in all cases). So it cannot be sure the operation it did is "valid" until checking the key after the operation. As a resultat such operation can only be read-only or rollbackable. This reduce the utility of the "same value == same tracked set" a lot. So it seems simpler to drop the double write and to update the documentation to highlight that this file does not garantee race-free operation. As a result the "key" is demoted to a "hint". Documentation is updated accordingly. Differential Revision: https://phab.mercurial-scm.org/D12201

tracked-file: rename the format option to use `use-` This is more consistent with the other options. Differential Revision: https://phab.mercurial-scm.org/D12200

tracked-key: update the requirement value We renamed the config option but we forgot to change the actual value… Differential Revision: https://phab.mercurial-scm.org/D12199

tracked-key: make it possible to upgrade to and downgrade from the feature This seems rather important if we want people to start using it. Differential Revision: https://phab.mercurial-scm.org/D12198

obsolete: don't use os.stat in repo.obsstore.__nonzero__ if it's static HTTP If a repo is accessed via static HTTP, then we obviously can't use os.stat() to just peek at the file size. Let's download the entire file to check its size. Yes, this feels wasteful, but: 1. If we're cloning or pulling a repo from a static HTTP server, we need the contents of the obsstore anyway. 2. Implementing statichttpvfs.stat() that uses HEAD will result in one more request to a static-only HTTP server, which is already slow. Also parsing a response to a HEAD request to construct os.stat_result is pretty hacky. There's also a question of the remote server properly supporting HEAD method and reporting at least file size. 3. Implementing statichttpvfs.stat() that uses GET is pretty much the same thing as we do here, except we can't even cache the response easily, unlike simply accessing obsstore._data, which is @propertycache'd. Importing statichttprepo locally to avoid circular import. See also: 4507bc001365 and commit message of f8f2ecdde4b5. Differential Revision: https://phab.mercurial-scm.org/D12195

filemerge: make `_maketempfiles()` more reusable `_maketempfiles()` is very specialized for its current use. I hope to use it also when creating temporary files for input for tools that do partial conflict resolution. That'll be possible if the function is more generic. Instead of passing in two contexts (for "other" and "base") and an optional path (for "local"), let's pass a single list of files to make backups for. Even if we don't end up using for partial conflict resolution, this is still a simplification (but I do have a WIP patch for partial conflict resolution and it is able to benefit from this). Differential Revision: https://phab.mercurial-scm.org/D12193

filemerge: reduce some duplication in `_maketempfiles()` The two callers of the local `maketempfrompath()` function used the returned file object in the same way. We can reduce duplication by moving that code into the function. Differential Revision: https://phab.mercurial-scm.org/D12192

filemerge: use leverage `util.readfile()` in `_maketempfiles()` Differential Revision: https://phab.mercurial-scm.org/D12191

filemerge: move removal of `.orig` extension on temp file close to context The place where the `.orig` extension is removed in `_maketempfiles()` doesn't make it clear that it's the backup path, which is why we have a comment in the code explaining it. Let's instead move it out of the function and close to where we get it from `backup.path()`, so that becomes clear. Differential Revision: https://phab.mercurial-scm.org/D12190

filemerge: remove `uselocalpath` argument from `_maketempfiles()` The `localpath` argument is unused if `uselocalpath` is false, so we can use `None` as a sentinel value for the variable instead and remove extra `uselocalpath` argument. That's not much of a win, but it's a small step towards further improvements. Differential Revision: https://phab.mercurial-scm.org/D12189

filemerge: remove an unnecessary join with absolute path The `backup` path is now always absolute, so we don't need to join it with the working copy path. Differential Revision: https://phab.mercurial-scm.org/D12188

filemerge: when using in-memory merge, always put backup files in temp dir Before calling a merge tool, we create a backup of the local side of the merge. That file can be put in the working copy or in a temporary directory, depending on the user's config. When we're merging in memory, we don't want to write to the actual, on-disk working copy, so we write the file to the in-memory working copy instead. However, since we don't support external merge tools with in-memory merge, it makes no difference where the file is actually stored (and if we ever do add support for external merge tools, then the file clearly can't live in the in-memory working-copy object anyway). So, since it doesn't matter where the file is stored, we can simplify by always putting them in the system's temp directory. Differential Revision: https://phab.mercurial-scm.org/D12187

filemerge: remove unused `repo` argument from `_maketempfiles()` I missed this in D12171. Differential Revision: https://phab.mercurial-scm.org/D12194

upgrade: make dirstate-v2 format variant compatible with share This only affects the dirstate and is safe to upgrade in the share. Differential Revision: https://phab.mercurial-scm.org/D12197

upgrade: prepare code (and output) for the idea of upgrading share This would work for a subset of action only. Our first target is dirstate-v2. Differential Revision: https://phab.mercurial-scm.org/D12196

sparse: add timing block for parsing sparse configs This was showing up in an operation I was doing today, and I'd like to be able to get trace spans for it instead of just profiler samples. Differential Revision: https://phab.mercurial-scm.org/D12186

narrowspec: add timing block for validating narrowspec This was showing up in an operation I was doing today, and I'd like to be able to get trace spans for it instead of just profiler samples. Differential Revision: https://phab.mercurial-scm.org/D12185

Added signature for changeset dcec16e799dd

Added tag 6.0.3 for changeset dcec16e799dd

status: fix hg status race against file deletion Differential Revision: https://phab.mercurial-scm.org/D12202

branching: merge with stable

filemerge: put temporary files in single temp dir by default The feature introduced in D2888 seems like a pure improvement to me. It makes the names' of temporary file easier to read. Let's have it always enabled. I also removed the config option for the path prefix because it doesn't seem useful. I asked Kyle (the author of the feature) about it and he couldn't think of a reason to keep it. I suspect it was just that we to have a config to turn it on/off while it was experimental, so it might as well be a configurable prefix then. Differential Revision: https://phab.mercurial-scm.org/D12171

dirstate-tracked-key: update the config value to match latest discussion Special cases are not special enough, we align the option name on the other. The `version` value is undocumented for now as it can only have a single value. It is supported in the code to properly detect and abort if more value are introduced in the future value. Differential Revision: https://phab.mercurial-scm.org/D12184

rhg: signal when falling back in logs We use the `trace` level for the actual message because it can get really busy. Differential Revision: https://phab.mercurial-scm.org/D11814

revlog: register changelogv2 C implementation in parsers This allows Python code to make use of the C implementation of the changelogv2 base operations when the C extensions are enabled. The `format_version` values are now shared between the C and Python sides, avoiding an additional translation for the selection of the format version to use. Differential Revision: https://phab.mercurial-scm.org/D12179

revlog: implement changelogv2 packing and unpacking in C This introduces a C implementation of changelogv2 records packing and unpacking operations matching the pure Python counterpart, similarly to what we already have for revlogv1 and revlogv2. This is also necessary to access changelogv2 record fields from future Rust code without going through the Python part, which would annihilate any performance benefit. Differential Revision: https://phab.mercurial-scm.org/D12178

revlog: memset whole record instead of dealing with padding This is less error-prone. Differential Revision: https://phab.mercurial-scm.org/D12177

revlog: do not compute node location by hand in index_invalidate_added The node is not guaranteed to be at the same location all the time (e.g: changelog v2), so let's use the official existing API to get that value. Differential Revision: https://phab.mercurial-scm.org/D12176

rank: compute property incrementally This replaces the naive rank computation with a more efficient incremental method, avoiding computing the whole ancestor set when possible. Differential Revision: https://phab.mercurial-scm.org/D12143

rank: add test with golden values This adds a regression test for the computation of the rank, using the current values computed with the naive algorithm as the "golden" reference. Differential Revision: https://phab.mercurial-scm.org/D12142

rank: add minimal test This adds a small test checking the rank computation in the case of a merge. Differential Revision: https://phab.mercurial-scm.org/D12141

rank: add context and template keyword This makes the stored rank property accessible, to be expanded and printed. Differential Revision: https://phab.mercurial-scm.org/D12140

rank: naive rank property computation and retrieval This stores the rank (size of the ancestor set of a revision, including itself) in a changelog field and allows this property to be retrieved. This new property is used as part of stable-range computations, which will be introduced later on. The value is computed in a naive way from the definition of the rank. This will be replaced by a more efficient version subsequently. Differential Revision: https://phab.mercurial-scm.org/D12139

filemerge: move check for identical sides out of filemerge() `filemerge.filemerge()` returns `None` if no merge was necessary because the two sides were identical. I don't think it should be that function's responsibility to handle that case; we should ideally not even call `filemerge.filemerge()` if the two inputs identical. This patch therefore moves the check out to the caller (`mergestate.py`). The largefiles test changed because we now notice that the two sides made the same change, so we don't consider it a merge. Also note that the new message better matches the line above it in the test output. Differential Revision: https://phab.mercurial-scm.org/D12154

mergestate: use an early return for trivial merges `filemerge.filemerge()` returns `None` if no merge was needed because the two sides were identical. I'd like to move that to the caller. This is a little refactoring to prepare for that. Differential Revision: https://phab.mercurial-scm.org/D12153

simplemerge: remove now-unused arguments The `ui` argument is no longer needed now that we don't handle `--print` in the library code. Differential Revision: https://phab.mercurial-scm.org/D12151

merge: fix documented order of `labels` argument Differential Revision: https://phab.mercurial-scm.org/D12150

filemerge: when not keeping premerge, don't write markers to context When premerge is enabled (as it is for non-binary inputs by default) and the markers are not kept, we currently still write it to the output context and then restore the previous content right after. With the refactoring in the previous patch, we can easily avoid that step and instead write the output in the opposite case (i.e. when it's successful or when the markers are supposed to be kept). Differential Revision: https://phab.mercurial-scm.org/D12149

simplemerge: move printing of merge result to extension The `mercurial.simplemerge` module still has some command-lines processing that doesn't belong in such a low-level module. This patch moves the handling of `hg simplemerge --print` to the extension by having `mercurial.simplemerge.simplemerge()` return the merged text. Differential Revision: https://phab.mercurial-scm.org/D12148

simplemerge: remove code for checking binary input now that callers do it The callers now do the checking for binary inputs and handle warnings and/or errors, so we can remove that code from the low-level `simplemerge` module now. After this patch we just raise an error unless the caller told us to allow binary inputs. Differential Revision: https://phab.mercurial-scm.org/D12169

simplemerge: let filemerge check for binary inputs This is similar to the previous patch, but here we put a specialized copy of `simplemerge._verifytext()` in the the `filemerge` module instead. Differential Revision: https://phab.mercurial-scm.org/D12147

simplemerge: let extension check for binary inputs (unless `--text`) The `simplemerge` module does too much UI-related things. This patch puts a specialized copy of `simplemerge._verifytext()` in the simplemerge extension. Differential Revision: https://phab.mercurial-scm.org/D12146

simplemerge: store input data in MergeInput I want to be able to update the file contents for support for partial conflict resolution. Having the text available (and later settable) on `MergeInput` will enable that. Differential Revision: https://phab.mercurial-scm.org/D12144

simplemerge: convert MergeInput to regular, non-attr.ib class I'm about to add more logic and state to it. Differential Revision: https://phab.mercurial-scm.org/D12145

simplemerge: replace `**opts` passed to `simplemerge()` by keyword arguments The `simplemerge` module is library code; it should not get an unmodified `opts` dict from the `simplemerge` extension. Differential Revision: https://phab.mercurial-scm.org/D12152

automation: use gp3 volume type This is a newer volume type. It offers a better baseline of performance by default and is ~10% cheaper. It doesn't offer burst credits. But its performance is guaranteed, unlike gp2, which had wonky performance behavior. Differential Revision: https://phab.mercurial-scm.org/D12132

automation: use m6i instances This instance type is much, much faster than t3 and can perform tasks much quicker. Differential Revision: https://phab.mercurial-scm.org/D12131

narrow: add support for merging add and remove outside of the tracked set This extend f1eb77dceb36 to test and support the remaining action. Or, at least, the simple incarnation of them. Differential Revision: https://phab.mercurial-scm.org/D12181

status: prefer relative paths in Rust code … when the repository root is under the current directory, so the kernel needs to traverse fewer directory in every call to `read_dir` or `symlink_metadata`. Better yet would be to use libc functions like `openat` and `fstatat` to remove such repeated traversals entirely, but the standard library does not provide APIs based on those. Maybe with a crate like https://crates.io/crates/openat instead? Benchmarks of `rhg status` show that this patch is neutral in some configurations, and makes the command up to ~20% faster in others. Below is semi-arbitrary subset of results. The four numeric columns are: time (in seconds) with this changeset’s parent, time with this changeset, time difference (negative is better), time ratio (less than 1 is better). ``` mercurial-dirstate-v1 | default-plain-clean.no-iu.pbr | 0.0061 -> 0.0059: -0.0002 (0.97) mercurial-dirstate-v2 | default-plain-clean.no-iu.pbr | 0.0029 -> 0.0028: -0.0001 (0.97) mozilla-dirstate-v1 | default-plain-clean.no-iu.pbr | 0.2110 -> 0.2102: -0.0007 (1.00) mozilla-dirstate-v2 | default-copies-clean.ignored.pbr | 0.0489 -> 0.0401: -0.0088 (0.82) mozilla-dirstate-v2 | default-copies-clean.no-iu.pbr | 0.0479 -> 0.0393: -0.0085 (0.82) mozilla-dirstate-v2 | default-copies-large.all.pbr | 0.1262 -> 0.1210: -0.0051 (0.96) mozilla-dirstate-v2 | default-copies-small.ignored-unknown.pbr | 0.1262 -> 0.1200: -0.0062 (0.95) mozilla-dirstate-v2 | default-copies-small.ignored.pbr | 0.0536 -> 0.0417: -0.0119 (0.78) mozilla-dirstate-v2 | default-copies-small.no-iu.pbr | 0.0482 -> 0.0393: -0.0089 (0.81) mozilla-dirstate-v2 | default-plain-clean.ignored.pbr | 0.0518 -> 0.0402: -0.0116 (0.78) mozilla-dirstate-v2 | default-plain-clean.no-iu.pbr | 0.0481 -> 0.0392: -0.0088 (0.82) mozilla-dirstate-v2 | default-plain-large.all.pbr | 0.1271 -> 0.1218: -0.0052 (0.96) mozilla-dirstate-v2 | default-plain-small.ignored-unknown.pbr | 0.1225 -> 0.1202: -0.0022 (0.98) mozilla-dirstate-v2 | default-plain-small.ignored.pbr | 0.0510 -> 0.0418: -0.0092 (0.82) mozilla-dirstate-v2 | default-plain-small.no-iu.pbr | 0.0480 -> 0.0394: -0.0086 (0.82) netbeans-dirstate-v1 | default-plain-clean.no-iu.pbr | 0.1442 -> 0.1422: -0.0020 (0.99) netbeans-dirstate-v2 | default-plain-clean.no-iu.pbr | 0.0325 -> 0.0282: -0.0043 (0.87) ``` Differential Revision: https://phab.mercurial-scm.org/D12175

rust: remove unused `StatusError::IO` enum variant All `io::Error` cases are now handled through PatternFileWarning or BadMatch Differential Revision: https://phab.mercurial-scm.org/D12174

copyright: it's 2022 Differential Revision: https://phab.mercurial-scm.org/D12183

hghave: update rustfmt criterion In 9ebc10ad4a04 I updated `rustfmt` without touching hghave, which means that the CI has been skipping the format test ever since. Thankfully, only one offending line exists in the code that's been introduced since. Differential Revision: https://phab.mercurial-scm.org/D12180

dirstate-v2: fix infinite loop in pure packer Due to the naive approach to path relative-ness, some tree shapes like the one introduced in the associated test could result in the packer going into an endless loop which allocated new `Node` objects endlessly until the process was killed by Linux's OOM killer. Differential Revision: https://phab.mercurial-scm.org/D12170

contrib: refer to RHEL and derivatives instead of CentOS specifically Differential Revision: https://phab.mercurial-scm.org/D12182

packaging: replace centos8 by rockylinux 8 Differential Revision: https://phab.mercurial-scm.org/D12172

rust: fix code formatting Differential Revision: https://phab.mercurial-scm.org/D12173

sparse: rework debugsparse's interface hg debugsparse supports arguments like --include, similar to `hg tracked --addinclude` or `hg log --include`. But in `hg debugsparse`, the pattern is not an argument of the flag, instead the patterns are the anonymous command line arguments. Not only is this surprising, it makes it impossible to use --include and --exclude in the same invocation, or --reset --exclude. So I propose making debugsparse making --include, --exclude take an argument, and rejecting anonymous command line arguments, as well as allowing mixing several of these flags in one invocations. Differential Revision: https://phab.mercurial-scm.org/D12155

color: don't infer vt status from TERM on Windows Previously, it was assumed that Windows environments with "xterm" in the TERM environment variable meant that either "virtual terminal mode" was already enabled, or that we are running in an environment that didn't need a "virtual terminal mode" (such as mintty, that interprets ANSI sequences itself). However, modern Cygwin and MSYS2 set TERM=xterm when using the Command Prompt as the terminal, which needs "virtual terminal mode" to be manually enabled. However, due to (issue6640), the vtmode wasn't being enabled. This patch ensures that we always try to enable vtmode on Windows regardless of the state of TERM, so that: * ANSI-based colors work in modern Cygwin/MSYS2 (with Command Prompt), and * The vtmode is unnecessarily set when running in a different terminal such as mintty, but it is simply redundant and doesn't appear to have ill effects. Differential Revision: https://phab.mercurial-scm.org/D12158

rhg: enable `rhg status` by default, without config or env opt-in The full test suite now passes with `hg` pointing to rhg. Differential Revision: https://phab.mercurial-scm.org/D12162

rhg: Colorize `rhg status` output when appropriate Differential Revision: https://phab.mercurial-scm.org/D12168

rhg: Add support for colored output The same "label" system is used as in Python code Differential Revision: https://phab.mercurial-scm.org/D12167

rhg: Add parsing for the --color global CLI argument Differential Revision: https://phab.mercurial-scm.org/D12166

rhg: Make Ui::new falliable, add Ui::new_infallible This allows propagating color configuration errors Differential Revision: https://phab.mercurial-scm.org/D12165

rhg: Pass a &Config to Ui::new When a Ui object is needed to print errors about configuration-loading errors, an empty (default) configuration is used. Differential Revision: https://phab.mercurial-scm.org/D12164

rhg: Add support for HGPLAINEXPECT Differential Revision: https://phab.mercurial-scm.org/D12163

merge: with stable

help: make the help for `safe-mismatch` real section The extra new lines mean the minirst engine were not seeing them as section… meaning they could not be addressed directly. Differential Revision: https://phab.mercurial-scm.org/D12161

help: avoid repeated value in the safe-mistmatch help This was copy pasted and overlooked when it got in. Differential Revision: https://phab.mercurial-scm.org/D12160

help: add missing `.` in config reference The `hg help` command is still confused at the number of dots, but at least the reference is correct. Differential Revision: https://phab.mercurial-scm.org/D12159

scmutil: obsrevs is already a frozenset Differential Revision: https://phab.mercurial-scm.org/D12157

obsolete: always return frozensets from obsolete.getrevs() getrevs function already returns an empty frozenset when there is no obsstore, but let's make sure to return a frozenset in any case. This makes it possible to use the result of this function as a dict key or provide it to hash() built-in function without any conversions. Differential Revision: https://phab.mercurial-scm.org/D12156

revlog: split revlog v1 and revlog v2 handling Explicitly splitting their fields packing and unpacking makes it easier to extend the existing C implemenation to handle the new changelog format, whose fields and offsets are not simply a superset of the revlog. Differential Revision: https://phab.mercurial-scm.org/D12137

revlog: extract entry byte offsets into named constants Labelling the fields pointed by the given offsets shared by revlog v1 and v2. Differential Revision: https://phab.mercurial-scm.org/D12136

doc: inspect.getargspec has been removed in Python 3.11 Fixed by dropping the inspection introduced in cdda48c93676. The 2nd "reporter" parameter to docutils.languages.get_language has been available since 0.8 more than 10 years ago. Reported for Fedora on https://bugzilla.redhat.com/show_bug.cgi?id=2022252#c2 .

branchmap: split a long condition in branchcache.validfor(), add comments Differential Revision: https://phab.mercurial-scm.org/D12138

branchmap: don't add branch entries if there are no heads We definitely don't want any empty entries to be present in repo.branchmap() just for the sake of not breaking test-notify.t. No test changes required because the previous patch made notify extension to not raise any tracebacks in case of RepoLookupErrors. Differential Revision: https://phab.mercurial-scm.org/D12135

notify: don't produce errors if a revision is not found Notify extension has a way to only subscribe to a specific revset, such as "branch(foo)". Before this patch, when there was no branch with that name, it would produce a traceback saying "unknown revision: foo". With this patch it would no longer do that, and instead it'll assume there are no revisions that match this revset. I think this patch is an improvement in general, but there's a reason I'm sending it now. test-notify.t has a test case where it obsoletes the only revision on a branch, and previously that wouldn't produce any complications, because head computation wasn't obsolescence-aware. Now if the only revision on a branch is obsolete, repo should not see that branch at all. That branch will still be present in branchcache (with an empty list of revisions) until the next patch. Differential Revision: https://phab.mercurial-scm.org/D12134

narrow: allow merging non-conflicting change outside of the narrow spec We use the mergestate to carry information about these merge action and reprocess them at commit time to apply the necessary update. The dirstate itself is never affected and remains "pure", with content only in the narrow-spec. This file involved in such merge are therefor not listed in `hg status`. The current testing is based on a modification of the previous testing, that refused to do such merges. As a result it is a bit simple and more extensive code and testing testing will have to be introduced later. I am planning to do this extra testing, soon. In addition, this only works for flat manifest. Support for tree manifest will need more work. I am not currently planning to do this work. Differential Revision: https://phab.mercurial-scm.org/D12119

merge-actions: add some information about the "changes" the action do This will be useful when processing merges action outside of the narrow-spec. "support" outside of narrow file on commit Differential Revision: https://phab.mercurial-scm.org/D12118

merge-actions: have an attribute for narrow safetiness This allow the core doing narrow filtering to process action without explicitely listing all possible actions. This is important to make the actions system more flexible in the future. Differential Revision: https://phab.mercurial-scm.org/D12117

merge-actions: add an explicite "no_op" attribute This make the MergeAction smarter and able to describe themself. This is useful to help introducing more MergeAction object that better the complexity of the situation. Differential Revision: https://phab.mercurial-scm.org/D12116

merge-actions: gather all created action into a set This is mostly to demonstrate we can do this before we start adding more specialized set. Differential Revision: https://phab.mercurial-scm.org/D12115

merge-actions: make merge action a full featured object This open the way for having "smarter" value as action, making the usage code simpler and more flexible. We have to explicitly use __bytes__ call in a couple of place because Python2… Differential Revision: https://phab.mercurial-scm.org/D12114

convert: use the merge action constant The constant exists, lets use them. Otherwise we cannot make these constant more powerful. Differential Revision: https://phab.mercurial-scm.org/D12113

large-file: use the merge action constant The constant exists, lets use them. Otherwise we cannot make these constant more powerful. Differential Revision: https://phab.mercurial-scm.org/D12112

merge: stop using merge action for pathconflict option This is not the b'r' you are looking for. Differential Revision: https://phab.mercurial-scm.org/D12111

merge: break up a not-so-one-liner for readability Differential Revision: https://phab.mercurial-scm.org/D12110

merge: break up a not-so-one-liner for readability Differential Revision: https://phab.mercurial-scm.org/D12109

merge: break up a not-so-one-liner for readability Differential Revision: https://phab.mercurial-scm.org/D12108

merge: break up a not-so-one-liner for readability Differential Revision: https://phab.mercurial-scm.org/D12107

merge: break up two not-so-one-liner for extra readability Differential Revision: https://phab.mercurial-scm.org/D12106

merge: break up a not-so-one-liner for readability Differential Revision: https://phab.mercurial-scm.org/D12105

merge: break up a not-so-one-liner for readability (even if not fully satisfied this time) Differential Revision: https://phab.mercurial-scm.org/D12104

merge: break up a not-so-one-liner for readability Differential Revision: https://phab.mercurial-scm.org/D12103

merge: break up a not-so-one-liner for readability Differential Revision: https://phab.mercurial-scm.org/D12102

merge: break up a not-so-one-liner for readability Differential Revision: https://phab.mercurial-scm.org/D12101

phabricator-refresh: add a magic value to skip it in the CI When maintaining (light) fork, it is useful to be able to disable this. Differential Revision: https://phab.mercurial-scm.org/D12133

interfaces: add missing caches kwarg of localrepo.updatecaches() Added to localrepo: in 1337bfaa88ca, added to the interface: now. Differential Revision: https://phab.mercurial-scm.org/D12092

obsolete: don't import from .node twice See d55b71393907. Differential Revision: https://phab.mercurial-scm.org/D12121

dagop: don't import nullrev from .node twice See 59fa3890d40a. Differential Revision: https://phab.mercurial-scm.org/D12120

stream-clone: allow to change persistent-nodemap format during stream clone Persistent nodemap affect the store format. However it is fairly isolated and fast to generate locally. So not making it a fixed part of the stream clone is useful. This allow clients without persistent-nodemap support (default for client without Rust enabled, or simply older client). So it make it possible to enable persistent nodemap on client, where it can provide a massive boost. without too much consequence. To do so, we stop using it in the advertisement requirements for streaming and let the client add/remove the necessary file depending of its configuration. We still send the files as it seems like a small save to not regenerate them. In addition, the way we match them will overlap with the changelog-v2/revlog-v2 so we can't simply skip the associated patterns. Differential Revision: https://phab.mercurial-scm.org/D12096

stream-clone: add 5000 changesets to test-clone-stream-format This make sure the changelog is not inlined. We needs this to test coming behavior change around persistent-nodemap. Differential Revision: https://phab.mercurial-scm.org/D12095

debugbuilddag: add a flag to allow running it from a non-empty repository Allow that by default seems "dangerous", but having a flag to make it possible will be useful to help building some repository incrementally. The newly introduced support is basic, but already useful. Differential Revision: https://phab.mercurial-scm.org/D12094

persistent-nodemap: properly delete all nodemap files on downgrade Previously, we were only deleting the docket file… Differential Revision: https://phab.mercurial-scm.org/D12093

destutil: if wdp is obsolete, update to the closest non-obsolete ancestor As the original comments suggest, using prune as a model here was an existing idea, and now this patch implements it. I think it would be even better to do what solveobswdp() from evolve does (in short, it allows to update to a successor of the closest ancestor even if that ancestor is obsolete), but that is outside of this series' scope. Differential Revision: https://phab.mercurial-scm.org/D12099

discovery: port _postprocessobsolete() changes from evolve, add tests Let's make this function obsolescence-aware as well. Now we check that revisions are not obsolete first, and only then check that they are going to be common after exchange. The tests make sure that rewriting changesets doesn't confuse head computation. They rely on experimental.single-head-per-branch feature to show that remote correctly sees the expected result (i.e. either aborts the push or allows it to succeed). They are ported from evolve as well. Differential Revision: https://phab.mercurial-scm.org/D12098

branchmap: skip obsolete revisions while computing heads It's time to make this part of core Mercurial obsolescence-aware. Not considering obsolete revisions when computing heads is clearly what Mercurial should do. But there are a couple of small issues: - Let's say tip of the repo is obsolete. There are two ways of finding tiprev for branchcache (both are in use): looking at input data for update() and looking at computed heads after update(). Previously, repo tip would be tiprev of the branchcache. With this patch, an obsolete revision can no longer be tiprev. And depending on what way we use for finding tiprev (input data vs computed heads) we'll get a different result. This is relevant when recomputing cache key from cache contents, and may lead to updating cache for obsolete revisions multiple times (not from scratch, because it still would be considered valid for a subset of revisions in the repo). - If all commits on a branch are obsolete, the branchcache will include that branch, but the list of heads will be empty (that's why there's now `if not heads` when recomputing tiprev/tipnode from cache contents). Having an entry for every branch is currently required for notify extension (and test-notify.t to pass), because notify doesn't handle revsets in its subscription config very well and will throw an error if e.g. a branch doesn't exist. - Cloning static HTTP repos may try to stat() a non-existent obsstore file. The issue is that we now care about obsolescence during clone, but statichttpvfs doesn't implement a stat method, so a regular vfs.stat() is used, and it assumes that file is local and calls os.stat(). During a clone, we're trying to stat() .hg/store/obsstore, but in static HTTP case we provide a literal URL to the obsstore file on the remote as if it were a local file path. On windows it actually results in a failure in test-static-http.t. The first issue is going to be addressed in a series dedicated to making sure branchcache is properly and timely written on disk (it wasn't perfect even before this patch, but there aren't enough tests to demonstrate that). The second issue will be addressed in a future patch for notify extension that will make it not raise an exception if a branch doesn't exist. And the third one was partially addressed in the previous patch in this series and will be properly fixed in a future patch when this series is accepted. filteredhash() grows a keyword argument to make sure that branchcache is also invalidated when there are new obsolete revisions in its repo view. This way the on-disk cache format is unchanged and compatible between versions (although it will obviously be recomputed when switching versions before/after this patch and the repo has obsolete revisions). There's one test that uses plain `hg up` without arguments while updated to a pruned commit. To make this test pass, simply return current working directory parent. Later in this series this code will be replaced by what prune command does: updating to the closest non-obsolete ancestor. Test changes: test-branch-change.t: update branch head and cache update message. The head of default listed in hg heads is changed because revision 2 was rewritten as 7, and 1 is the closest ancestor on the same branch, so it's the head of default now. The cache invalidation message appears now because of the cache hash change, since we're now accounting for obsolete revisions. Here's some context: "served.hidden" repo filter means everything is visible (no filtered revisions), so before this series branch2-served.hidden file would not contain any cache hash, only revnum and node. Now it also has a hash when there are obsolete changesets in the repo. The command that the message appears for is changing branch of 5 and 6, which are now obsolete, so the cache hash changes. In general, when cache is simply out-of-date, it can be updated using the old version as a base. But if cache hash differs, then the cache for that particular repo filter is recomputed (at least with the current implementation). This is what happens here. test-obsmarker-template.t: the pull reports 2 heads changed, but after that the repo correctly sees only 1. The new message could be better, but it's still an improvement over the previous one where hg pull suggested merging with an obsolete revision. test-obsolete.t: we can see these revisions in hg log --hidden, but they shouldn't be considered heads even with --hidden. test-rebase-obsolete{,2}.t: there were new heads created previously after making new orphan changesets, but they weren't detected. Now we are properly detecting and reporting them. test-rebase-obsolete4.t: there's only one head now because the other head is pruned and was falsely reported before. test-static-http.t: add obsstore to the list of requested files. This file doesn't exist on the remotes, but clients want it anyway (they get 404). This is fine, because there are other nonexistent files that clients request, like .hg/bookmarks or .hg/cache/tags2-served. Differential Revision: https://phab.mercurial-scm.org/D12097

obsolete: make sure windows tests pass when stat() is given a URL This is a temporary fix for an issue that's only visible on windows, but exists on other platforms as well. The issue is if we're trying to use obsstore from a remote peer and that peer is a static HTTP repo, vfs tries to os.stat() a remote file using a URL to that file (e.g. http://localhost/repo/.hg/store/obsstore). The next patch in this series makes branchcache obsolescence-aware, so in certain situations exchange process will try os.stat()ing a URL. On windows this will produce an OSError that is not ENOENT, but EINVAL instead (because of `:` symbol, for example). Differential Revision: https://phab.mercurial-scm.org/D12100