obshistory: add _originmarkers() function to be used for obslog --origin

evolve: rename `evolve_source_*` to `divergence_source_*` This makes it clearer that this comes from divergence resolution.

evolve: mark new hash-salting extras keys as uninteresting for effect flag The user doesn't care if the `evolve_source_local`/`evolve_source_other` values in the extras changed, so this patch adds them to the blacklist that already exists for that purpose. This should also fix the flakiness we've seen in some tests. The flakiness seemed to stem from `mercurial.obsutil.geteffectflag()` not sorting extras keys. That means we ran into it relatively often for the `evolve_source_local`/`evolve_source_other` keys because there were two of them. That sorting bug should be fixed upstream.

evolve: handle relocation during divergence resolution producing no changes When resolving divergence and the two divergent commits have different parents, we start by rebasing one of them to have the same parent as the other. That step can result in no changes to commit. When it does, we would crash with a TypeError before this patch. This patch fixes it by instead creating an empty commit in that scenario. The existing code then continues to attempt to merge it, which produces no changes, and no commit is created on top. The other side of the divergence is marked as successor as usual, so orphans from the side that became empty will be evolved to the right place (see test).

tests: show crash from divergence resolution resulting in empty commit When relocating a commit results in an empty commit (because the changes are already in the destination), we get a TypeError from evolve.

evolve: always create commit when resolving divergence When resolving content-divergence, the final commit we create may end up empty (which means that Mercurial won't even create it). We've had code for handling that in evolve ever since 41bf6c27a122 (evolve: stabilize now handle conflicting changeset, 2012-08-23). However, that resolved the issue by marking on the divergent commits as successor. As Pierre-Yves has pointed out (in other code reviews), we should instead be creating a new successor. So that's what this patch does. It does that by setting `ui.allowemptycommit` while creating the final commit. However, that is not enough, because we may end up creating the same nodeid as already existed (we'd then end up trying to mark the "new" commit a successor of itself). To solve that, we add some salt to the commit extras. That salt affects lots of tests.

tests: replace an X+X revset with something meaningful The change in aab2afc2bd82 seem to have been make too hastily.

tests: update section headers and comments in obslog-related tests

tests: use common obshistory_setup.sh script instead of manual setup This is now consistent with other test-evolve-obshistory*.t tests.

tests: remove "second diverged revision" from a phase divergence test It was basically testing the same thing twice with a different comment.

obshistory: omit duplicate notes of obsolescence operations When folding changesets, it's possible to provide a --note flag to e.g. fold or prune, which will store a note for that operation. But since there would be more than one marker for the fold operation and the provided note would be stored in all of them, _markersnotes() previously would return the same text message many times. We don't really want to see the same thing over and over, so let's make notes a set(). This will also make unrelated operations with the same note show the message only once, which is probably an improvement too.

changelog: adjust The changes only apply to newly created sqlite database. There are code pathes that drop existing ones, but that's a more complicated condition.

changelog: mention the sqlite fix

evolve: extract function for relocating divergent commit There was some duplication here, so let's extract to a function. I'm going to add a little bit more shared code next.

caches: preserve permissions of top-level .hg When using hg on a shared filesystem on UNIX, cache files normally inherit the permissions of the .hg directory. This is most commonly used to ensure everything is writable by all users. The sqlite3 cache files don't have a way to directly set the permission, so check if a special mode is necessary and try to apply them to newly created database files.

changelog: mention obsorigin template keyword

templatekw: add obsorigin keyword

obshistory: add predecessorsandmarkers()

obshistory: add a function to group predsets by fold-id

obshistory: add _predecessorsverb() similar to _successorsetverb()

obshistory: add a docstring to patchavailable()

tests: avoid referring to commits by hash in a few places This patch means that the next patch can be written without manual editing of the test cases.

tests: add more verification of output to divergence tests It's not obvious what the obsmarkers should look like in these tests and adding them highlights a bug that happens when commit we create to resolve the divergence doesn't get created because it would be empty. I also added a few graphlog calls to check that the divergence-resolution commit is in the right place (has right parentage).

tests: add debugobsolete calls before many obslog calls As requested by Pierre-Yves. The main reason is to catch superfluous creation of obsmarkers, as well as any future changes to how the obsmarkers get created.

stablerangecache: sanity check subranges Try to detect invalid conditions on insert as would result in deep recursions and final aborts much later. This has been observed on two different machines and the check makes it hopefully possible to find the origin of the problem.

obslog: use unfiltered repo for --no-graph implementation too repo.unfiltered() is used for graph version of obslog, because if we want to filter non-local changesets, we only have to provide --filternonlocal flag on the command line and _nodesandmarkers() will handle filtering. But that means for obslog without -f to work correctly we need to provide unfiltered repo to that function. With this patch both implementations of obslog (--graph and --no-graph) see the same repo.

tests: don't assume that echo will unescape \\

tests: avoid non-portable sed command sed -i is a GNU extension and the i command without newline as well. Just create the file from scratch with printf to side-step the issue.

tests: check for file existance in a portable way The precise output of ls is highly OS specific, so just use the simpler test command.

tests: deal with "ls" vs "ls -A" difference on 2BSD derived systems ls on 2BSD derived systems historically defaults to -A for root, so use it unconditionally for consistent results.

tests: silence potential warnings about uname -o The option is not supported on the BSDs, but also not relevant as it is only used to detect msys.

evolve: remove unused divergenceflag and splitflag of MultipleSuccessorsError

evolve: consider all obsolete ancestors when finding new parent for orphan Both the old and the new logic have in common that the orphan changeset is moved to the tipmost successor of one or more ancestors of the orphan changeset. The previous logic considered only the first non-pruned ancestor of the orphan changeset. The new logic considers all obsolete ancestors of the orphan changeset. The logic in _possibledestination() had to be changed to be consistent with the new logic.

py3: fix debugstablerangecache

topic: provide cache to successorssets() in _singlesuccessor() Since we're using obsutil.successorssets() in a loop, it may help performance to have cache.

evolve: provide cache to successorssets() in _singlesuccessor() Since we're using obsutil.successorssets() in a loop, it may help performance to have cache.

evolve: provide cache to successorssets() in _possibledestination() Since we're using obsutil.successorssets() in a loop, it may help performance to have cache.

obslog: support --all with --no-graph

obslog: support --filternonlocal with --no-graph

obshistory: restructure templates a bit, drop default verb ("pruned") We always have a verb provided to the templates now, this should be fine.

obshistory: prepare data for displaymarkers() in a separate function

obshistory: make patchavailable() handle predecessors too

tests: make sure not to test the same thing twice A block under this one already tests obslog --no-graph with -f. Right now there's no difference in output, but maybe some day obslog --no-graph will actually support --filternonlocal.

obshistory: rename precursors to predecessors in displayrevs()

obshistory: rename _debugobshistorydisplaymarker() to displaymarkers() Apart from the obvious change, the function now can show multiple markers, not just one.

obshistory: rename graph/revs-displaying functions Same idea as the previous patch, but with a "display" prefix.

obshistory: rename node/ctx-displaying functions Not just for debug, and yes we know it's obshistory-related.

branching: merge with stable

topic: use generationssubrel less directly, because it was recently renamed

auto-publish: issue the capabilities in all cases It seems better to message client that we support the capabilities in all cases.

packaging: add an entry to debian/changelog mentioning the previous change

packaging: bump minimum required version of mercurial in debian/control

[debian] package 9.3.1-3 for bionic ubuntu

tests: update output to match 1b8fd4af3318 in core CORE-TEST-OUTPUT-UPDATE: 1b8fd4af3318

manifest: now that there's only one legacy.py, exclude it After the last patch, this file is only included because of package_dir in setup.py (i.e. just once), so we can ignore it with just one clause. But package_dir starts with "./", so we need to add it to the pattern.

manifest: don't include .py files from hgext3rd/ twice They are already included by default because we've given package_dir to setup().

manifest: just include all .py .sh and .t files from tests/ recursively

manifest: make sure exclusions and prunes are last, because order matters

branching: merge 9.3.1 release into default

packaging: mark as developer version

Added tag 9.3.1 for changeset 8d955635cf45

packaging: prepare version 9.3.1

test-compat: merge mercurial-4.7 into mercurial-4.6

test-compat: merge mercurial-4.8 into mercurial-4.7

test-compat: merge mercurial-4.9 into mercurial-4.8

test-compat: merge mercurial-5.0 into mercurial-4.9

test-compat: merge mercurial-5.1 into mercurial-5.0

test-compat: merge mercurial-5.2 into mercurial-5.1

test-compat: merge stable into mercurial-5.2

topic: hg <= 4.8 compatibility for wrapping exchange._checkpublish() phase-divergent update to 1a81bbc94d45:

branching: merge with stable

topic: check that exchange._checkpublish() exists before wrapping it

head-checking: also test single head enforcement with topic They are an important part of why we want this.

head-checking: ignore obsoleted section when checking head creation locally Same as for the server side check, these do not need to be taken in account.

head-checking: ignore obsolete section when checking for single heads This overwrite the core code that enforce single heads to ignore obsolete section. That way older obsolete version of a branch no longer prevent the new version to be pushed.

head-checking: update changelog

head-checking: add test variant mixing branch and topics

head-checking: add test variants with topics

head-checking: backport fix from 7d5455b988ec See 7d5455b988ec for details. Having a version in the evolve repository allow us to check the behavior when topic is enabled. It will also become the base of further head checking fix and topic compatibility.

head-checking: filter out obsolete heads when checking for new heads This does not affect any tests right now, but this will be useful for the next changeset. This is also the start of a more aggressive removal of obsolete heads in places where it does not make sense to take them into account.

topics: fix auto-publish=abort with servers publishing bare branches Before this change, servers that automatically publish bare branches were detected as "non publishing" and the check for automatic publishing on push assumed nothing was to be published. Now we signal the server-side config to clients via a new capability so that they can adjust their behavior.

evolve: fixed some typos in stablerangecache LONG_MESSAGE (issue6279) phase-divergent update to bf37ba1c80ee

evolve: fixed some typos in stablerangecache LONG_MESSAGE, fix #6279

evolve: fixed some typos in stablerangecache LONG_MESSAGE, fix #6279

branching: merge stable in default

WIP: fix

WIP: rotate loop

WIP: test

build: make docs/conf.py compatible with py3 I don't even know what this file does, but it made our Debian build fail on py3 because it attempted to pass unicode to `mercurial.ui.setconfig()`.

obshistory: add _debugobshistorydisplaypredsandmarkers() and test it

obshistory: make patchavailable() handle predecessors too

templatekw: add obsorigin keyword

obshistory: implement predecessorsandmarkers() for obsorigin

fold: proposal for output improvement Listing other changeset we got folded with seems like a good idea. We need to think about the case where changesets we got folded with is filtered.

obslog: add display of precursor information to the default template This will still only be output when the user passes --precursors to the command (otherwise the data just won't be there). This addresses a common user request, where they actually would find more intuitive to use --precursors --no-successors.

obslog: add flags to control the output of precursors and successors With just this change, precursors are only shown in structured output.

obslog: extract display of a markers to its own function

tests: allow an empty merge commit to have tests pass without too many changes Not the greatest fix, but this allows us to have the same test case for all versions of Mercurial. Hash of 29 changes because instead of both parents only one is set in hg 5.3.

uncommit: use repo.setparents() instead of repo.dirstate.setparents() setparents() was added to localrepository class in 2012, and now we finally have some differences between calling repo.setparents() vs repo.dirstate.setparents(): _quick_access_* things, introduced in hg 5.3. This patch fixes potential quick-access desync problems where repo[b'.'] would be different from repo.dirstate.p1().

touch: use repo.setparents() instead of repo.dirstate.setparents() setparents() was added to localrepository class in 2012, and now we finally have some differences between calling repo.setparents() vs repo.dirstate.setparents(): _quick_access_* things, introduced in hg 5.3. This patch fixes potential quick-access desync problems where repo[b'.'] would be different from repo.dirstate.p1().

evolve: use repo.setparents() instead of repo.dirstate.setparents() setparents() was added to localrepository class in 2012, and now we finally have some differences between calling repo.setparents() vs repo.dirstate.setparents(): _quick_access_* things, introduced in hg 5.3. This patch fixes an actual quick-access desync problem where repo[b'.'] would be different from repo.dirstate.p1(), as demonstrated by test-evolve-phase-divergence.t when ran on hg 5.3rc0 and later. Modified _uncommitdirstate() too, because it is used in dirstatedance(), which sometimes follows repo.setparents() calls.

tests: demonstrate that evolve swaps parents of merge commits Tests currently pass, don't forget to change FIXME items.

ci: cache pytype-related files between runs

evolve: add pre-check logic for creation of phase divergence locally Changes in tests reflect the added behaviour.

evolve: add test to show that user can create phase-divergence locally After resolution of phase-divergence, user can locally create phase-divergence by rewriting the old bumped (obsolete now) changeset. Next patch will be adding the pre-check logic for creation of this phase-divergence.

evolve: remove unnecessary code since it's been covered already We don't need this logic any more since the case of "two divergent csets where one is the parent of other" has been handled correctly by the logic present in method _prepcontentdivresolution() This is how things works in method _prepcontentdivresolution() for our case i.e "content-divergence with parent-child relation": cset_b (content-divergent) | cset_a (content-divergent) | ~ Acc. to revision selection criteria: "divergent" = min_revision(cset_a, cset_b) So always "divergent" will be cset_a and "other" will be "cset_b" and resolution parent will be the successor of parent of cset_a Both the csets will be merged and resolution cset will be based on correct revision. The result could be wrong only in the case when "divergent" is cset_b which is not possible acc. to the current logic.

test: no need to look at full log

evolve: refactor content-divergence resolution logic > What is the case we are looking at? This is about refactoring the part of content-div resolution logic where it decides which cset should be relocated and where. > What is a "topologicial common ancestors" vs a "greatest common ancestors"? `tca` is an ancestor which we can decide/find by looking at the at graph visually for e.g ``` c3(*) c4(*) | | c2(x) c1(x) c5 | / \ | / c0 ``` (c5 is the successor of c2 and c1) now here, `tca` of c3 and c4 is: c0 `gca` of c3 and c4 is: c5 > What is the new top-level logic/behavior that makes it better? The old code had some unnecessary edge cases just because we were using `gca`, since it can point to a revision that is not a topological ancestor. For e.g see b779b40f996e Eventually, the code around this was getting messy unnecessarily. So I looked into it and found a simple and more robust approach. And in new code, it is simple and straightforward (and easy to understand), where we handle the following 4 cases when solving content-div: 1) when both are on the same parent => (no need to do anything special, and simply proceed) 2) both are on the different parent but a) `tca` is the parent of one of them or b) there is no non-obsolete revision between `tca` and one of the divergent cset. => (relocate one to the other side and proceed) 3) both are on different parents and `tca` is not the parent of any of them and there is at least one non-obsolete cset between tca and both the divergent cset i.e (tca::div1) and (tca::div2) both the ranges have at least one non-obs revision. => (this is the case which we don't handle yet, but the solution would be to prompt the user to choose an evolve destination.) 4) both are in the parent-child relation => (both are merged and new cset will be based on the successor of `tca`) Changes in test-evolve-issue5958.t demonstrate that new code also covered case4 because in a resolution of "two divergent csets with parent-child relation" there should be one cset as a result and no orphan revs (as you can see there was an orphan before this patch).

evolve: remove duplicated code See at line 500.

evolve: make necessary changes in a test to reflect a fix in next patch

evolve: add quotes around the action in msg of pre-checking

metaedit: use pre-checking before rewriting csets Changes in test file demonstrate the changed behaviour.

evolve: update the pre-check message for risk of orphans

evolve: add pre-check logic for creation of phase divergence locally Changes in tests reflect the added behaviour.

evolve: add test to show that user can create phase-divergence locally After resolution of phase-divergence, user can locally create phase-divergence by rewriting the old bumped (obsolete now) changeset. Next patch will be adding the pre-check logic for creation of this phase-divergence.

evolve: remove a check which is already done by rewriteutil.precheck() We have tests to check if fold is being performed on public cset. There is no changes in test files because testing that never touched the part this patch removes, and already caught in rewriteutil.precheck()

evolve: add pre-check logic for content-divergence in rewriteutil.precheck() For now, pre-check will abort if rewriting a rev create divergence (and config experimental.evolution.allowdivergence is not set to True). But this behaviour can be improved where instead of abort maybe we can confirm the user to either proceed with divergence or some other options depends on what command user is running. Changes in test file are used to overrides the pre-check for testing purpose (using experimental.evolution.allowdivergence=yes).

topic: make in-memory rebase preserve topic The override code thought that `__init__` would return the runtime instance, but it's actually the first argument to the function (the `self` argument), so the code had no effect at all before this patch. I think the bug only affected in-memory rebase because the in-working-copy rebase used the current topic, which was set correctly since 5156a67f66a6 (topics: update current topic to the topic of newly rebased commit (issue5551), 2017-06-29).

Temporary commit for time measurement of test file + pandoc

ci: Add CI job for building the documentation