evolvecmd: mark relocate() private

cleanup: replace .parents()[0] by .p1()

evolve: add more tests for --continue case when relocating "divergent"

evolve: use meaningful variable names for a return value

evolve: add logic to cover --continue case when relocating "divergent" Changes in test file demonstrate the added behaviour.

evolve: add test for `--continue` case when relocating "divergent" This test shows that for now we don't handle the continue case when relocating "divergent" cset hit merge conflict while resolving content-divergence. Next patch will be adding the support for --continue case.

evolve: use more specific key name to store in evolvestate Now, as we also relocate divergent cset if required, so let's use more specific key to track the relocation of "divergent" and "other".

evolve: move a test where it makes more sense

evolve: remove the unnecessary condition to check empty successors set I annotated the history of this "newer == [()]" condition and found that it was added in the initial stages of evolution project (more than 8 years ago) and there is no test where we get [()] as a successor set. So looks like "if not newer" is enough to check if it is empty. I also looked into obsutil.successorssets() implementation and I don't think it would give us [()] in any case.

evolve: delete redundant and incorrect code for updating dirstate In relocate() when we notice conflicts, we will try to fix up the dirstate by calling copies.duplicatecopies() and dirstatedance(). However, we had called _evolvemerge() just before that point and that has already fixed up the dirstate, so there's no need to do it again. Also, we did it incorrectly as the test case shows.

tests: demonstrate how evolve loses transitive copies when interrupted Let's say `b` has been renamed to `c` in commit A and then renamed to `d` in commit B. We now modify A in a way that will cause a conflict when we evolve B. `hg evolve` will now stop and let the user resolve the conflicts, but it will have lost the copy information in the working copy (should still be c->d).

evolve: use ui.configoverride() instead of ui.backupconfig() ui.configoverride() was added in hg 4.1. util.nullcontextmanager() was added in hg 4.4.

topic: use ui.configoverride() instead of ui.backupconfig() ui.configoverride() was added in hg 4.1. util.nullcontextmanager() was added in hg 4.4.

topic: use ui.configoverride() instead of ui.backupconfig() ui.configoverride() was added in hg 4.1. util.nullcontextmanager() was added in hg 4.4.

evolve: use ui.configoverride() instead of ui.backupconfig() ui.configoverride() was added in hg 4.1.

evolve: remove obsolete backup of allowemptycommit config We stopped setting the config in 88601e1cd5d8 (evolve: create new commit while resolving content-divergence, 2018-04-23), so there's no need to back it up.

evolve: remove dead initilization of "newnode" variable It's been dead since f9dad99a90d5 (evolve: create a new commit instead of amending one of the divergents, 2018-06-13).

evolve: fix a typo

changelog: mention the python 3 fix

changelog: mention the date fix for content divergence

branching: merge with stable

py3: fix several TypeError formatting bytes with an exception object

evolve: preserve date when resolving content-divergence (issue6202) This patch changes the implementation to do a 3 way merge. In case of actual conflict, it take the newest date from both sides of the content-divergence.

test: add test for content divergence involving a date update In this case, each side update the date.

test: add test for content divergence involving a date update In this case, one side update to an earlier date.

test: add test for content divergence involving a date update In this case, one side update to an older date.

test: highlight data handling when solving divergence The current handling of date is lacking, we add test to highlight this.

evolve: add tests for the case when resolution parent is ambiguous

evolve: abort if deciding resolution parent is ambiguous While solving content-divergence, it could be hard to decide which cset should be treated as a resolution parent if there are multiple successors of divergent cset. i.e split case. But we are planning to make some update in handling this split case: 1) consider highest one if all splitted csets are in a single topological branch 2) still need to decide if 1) doesn't apply

evolve: make sure divergence resolution doesn't undo changes (issue6203) Before this patch, in content-divergence resolution logic if resolution parent is not the parent of any of the two divergent changesets then it could undo some changes introduced by previous revs (while resolving stack of content-divergent changesets) as demonstrated by the test added in previous patch. To solve this, what this patch doing is: if divergent cset has obsolete parent with a successor then first resolve the "orphan" instability of divergent cset by relocating, then perform the content-divergence resolution. After this change in logic, I found that it's kind of more correct as reflected by the changes in tests/test-evolve-content-divergent-corner-cases.t where it prevented creating conflicts while merging. Changes in tests/test-evolve-content-divergent-stack.t demonstrate the fixed behaviour. Next patches will be covering the `evolve --continue` case for the relocation of "divergent" cset.

evolve: add test to demonstrate issue6203 Current logic of content-divergence resolution contains a bug that if resolution parent is not the parent of any of the two divergent changesets, it could undo some changes. I think a good solution for this is to first relocate the divergent cset to it's obsolete parent's successor if applicable and then perform the content-divergence resolution. Next patch will fix the issue.

state: there's no _unpack in struct Caught by pytype.

amend: abort if both --patch and --extract are used

test-compat: merge mercurial-4.6 into mercurial-4.5

ci: add job to build and archive the documentation

tests: update now that merge.graft() doesn't set two equal parents This is a consequence of https://phab.mercurial-scm.org/D7549. CORE-TEST-OUTPUT-UPDATE: 32d11a23c9cf

evolve: fix content-divergence resolution when p1 is null (issue6201) Before this fix, in cases like in the added test, the revision number -1 of the parent was misinterpreted to mean the tipmost revision instead of the null revision, causing the content-divergence resolution to fail.

obslog: make {node} the full node and leave shortening to template It should be up to the user to shorten the nodeid as much or as little as they like.

branching: merge with stable

compat: accept "success" argument to lock._afterlock callback The callback gets a new "success" argument since Mercurial commit 888bd39ed555 (lock: pass "success" boolean to _afterlock callbacks, 2019-11-19).

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

test-compat: run tests with -j more than 1 In 4.7 and below run-tests.py defaults to -j1. We want more speed, so we use the same mechanism to detect the available number of CPU threads as the newer versions of run-tests.py.

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: don't run the py3 test on the older branches

docs: add missing Figure 10 to sharing.rst The .svg file itself has been already added earlier.

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 stable into mercurial-5.1

heptapod-ci: add a script to map evolve branch to mercurial revision for tests map-hg-rev.sh can be tested using this command: ```sh for i in $(hg log -T '{branch}\n' | sort -u); do echo $i $(testlib/map-hg-rev.sh $i) done ```

pytype: ignore abc.ABCMeta metaclasses that don't work on py3 See 70d42e2ad9b4 in hg: We can remove this when we're Python 3-only, but for now it's just too awkward to deal with and it's harmless.

tests: remove extra whitespace before patch in obslog output The indent() template function was fixed in https://phab.mercurial-scm.org/D7432 so now we have less trailing whitespace in obslog output. CORE-TEST-OUTPUT-UPDATE: fa246ada356b

branching: merge with stable

changelog: add a couple of missing entry about the next feature release

changelog: fix entry in the wrong location That change was made on default, not stable.

changelog: add a wheelbarrow of missing entries

heptapod-ci: simpler CI We have a mercurial clone available in the docker image now. This will same some time to avoid a full Mercurial clone for each run.

heptapod-ci: run the test with color This make the result easier to read.

compat: compatibility for cl.nodemap.get vs cl.index.get_rev

compat: compatibility for pathuril.dirs vs util.dirs

obslog: phase divergence fix for 5c41bb482867 I forgot to pull the latest version before publishing the topic :-(

obslog: phase divergence fix for b135591bec1a I forgot to pull the latest version before publishing the topic :-(

obslog: make {descdiff} not be indented and leave that to the template Same idea as the previous patch and same explanation for changes to the test case.

obslog: make {patch} not be indented and leave that to the template This removes some unwanted indentation from the tests. In the first case, it's because indent() doesn't indent a blank line. In the second case, it's because we used templating with {patch} and had not requested indentation (but we still got it before this patch). Note that tests will only pass with a very recent Mercurial (one with https://phab.mercurial-scm.org/D7432 applied).

obslog: make {succnodes} be full hex nodes It should be up to the user to shorten the nodeid as much or as little as they like.

obslog: remove now-unused code for plain styling We now always use a non-plain formatter for the non-local-filtering code paths, so we can remove the code for plain styling. I couldn't figure out how to replace the fm.formatlist() by something less about formatting, so I've only removed the irrelevant arguments from those calls.

obslog: redefine default output as a template, unless -f was given This was easier than I expected to do without affecting any test output at all. `hg obslog -f` is pretty broken with templating. For example, it inserts a single "date" field, but it provides two values for it if the dates are different. That results in a traceback when running e.g. `hg obslog -f -Tfoo` with non-local obsmarkers with different dates. We need to figure out what we want to put in the template for the filtered case. I've just left it out for now.

obslog: add newline after all markers in non-graphlog to match graphlog I'd also be fine with doing the reverse if that's preferred, but I really want them to be the same so they can use the same template.

obslog: make reason for unavailable patch part of template We need the reason to be in the template in order to be able to templatize the whole output.

obslog: make content and description patches available to templater The code was repeatedly calling fm.write() with the same field ("patch" and "descdiff"). I think that led to the value constantly getting replaced, so when it was used in a template (as {patch} or {descdiff}), it would only get the last value, which was always an empty string. This patch fixes it by writing the full patch to a temporary buffer and then assigning the whole patch to the formatter field.

obslog: rename {desc-diff} to {descdiff} since the former is invalid

tests: demonstrate broken {patch} and {desc-diff} template keywords

topic: drop compat.successorssets(), it's not used The only place in topics that uses successorssets function is in evolvebits.py, and it's using that function from obsutil directly.

topic: drop compat.getmarkers() and use obsutil.getmarkers() directly The function in question was moved from obsolete to obsutil in a14e2e7f7d1f (hg 4.3). See also 6aff4bb3970d for hgext3rd/evolve.

cmdrewrite: avoid accessing scmutil.status fields by index Support for indexed access is going away in Mercurial. Accessing the fields by name is clearer anyway.

obslog: avoid using a formatter after calling end() on it I don't know if makes a difference, but it feels wrong to use it after calling end(). We can simply use the parent formatter in this case.

obslog: use plural name "effects" for list of all effects Same reasoning as the previous patch, but here the singular was used for the list instead. This patch also renames the variable that represents the list of effects to plural "effects" to reduce confusion.

obslog: use singular name "succnode" for each element of {succnodes} The name that we pass for formatlist() is the name of each element. After this patch, you'll write '{succnodes % "{succnode}"}' instead of the confusing '{succnodes % "{succnodes}"}' (where the two "succnodes" refer to different things. Users can write templates that are compatible across this change by using e.g. '{succnodes % "{if(succnode, succnode, succnodes)}"}'.

docs: add SVG figures for sharing.rst Generated by graphviz loosely based on .dot files produced by dotgraph extension.

docs: add some up-to-date output from push/pull commands Hopefully it's going to be helpful for users to better understand how evolve works.

docs: add two more amend commits to simulate temporary amend commits sharing.rst made reference to temporary amend commits and used them to demonstrate that hidden commits are not exchanged. Nowadays, evolve doesn't create such commits, but it still makes sense to show how they are handled during the exchange process. So let's add two more amend commits, one for each repo. This way the guide doesn't have to be updated too much, but doesn't lose this important detail of working with evolve. Unfortunately, this means that tons of hashes change, but it's better than to have figure 4 demonstrate absolutely nothing. Temporary amend commits were removed from test-sharing.t in 06844693bb21, but sharing.rst continued using them for demonstration purposes. It might've been better to replace at least some of the temporary amend commits by extra amends back then, but oh well.

docs: revision numbers are technically stable From `hg help glossary`: "Note that the revision number may be different in each clone of a repository." But cloning the same repo multiple times yields the same result, and that shows that revision numbers are stable.

tests: remove a repeated statement We looked at the review repo just 6 lines above.

docs: correct node hashes in sharing.rst based on test-sharing.t

tests: add shortlog of other repos to test-sharing.t The figures require showing multiple repos, so let's show them all, even if they look trivial.

tests: use |shortest in test-sharing.t because sharing.rst uses it

compat: compatibility for cl.nodemap.get vs cl.index.get_rev

compat: compatibility for pathuril.dirs vs util.dirs

branching: merge with stable

heptapod-ci: test with python3 too

heptapod-ci: add a --pure variant for the CI The end goal is to migrate all piece of the evolve CI currently en Jenkins.

heptapod-ci: let the number of CPU to be picked automatically We now have a larger, bigger machine to run the tests. Let's let run-tests.py pick the number of concurrent jobs.

heptapod-ci: point to our own docker images Lets step further into the future.

readme: update contribution guide We move to heptapod !

branching: merge with stable

docs: remove --all from hg evolve, since it's been the default for some time

docs: typo in user-guide.rst

obslog: don't overwrite {rev} keyword from changeset context Now that we have the whole changeset in the formatter context, there's no need to explicitly add {rev}. We should also deprecate the {shortdescription} keyword since it's no longer needed, but I don't know how to do that.

obslog: make changeset available to templater obslog templating currently has very limited support for templating the precursor changeset (when available). The only available keywords are {node}, {rev}, and {shortdescription}. This patch makes the entire changeset available to the templater by passing it to formatter.context(). That function has been available since Mercurial 4.2.

tests: simplify by pruning unwanted changesets a little earlier The pruned commits are not relevant to the test case, so let's clean them up at the end of the previous test case instead of part-way through the next one.

evolve: use more often seen metavariables in command synopsis strings Also use ellipsis when multiple options are acceptable.

topic: add more options to command synopsis string

debian: also build with python3 We build with Python 3 along with Python 2. The binary package will contain the extension for both python version. This is to ease transition until the mercurial package in Debian uses Python 3. The documentation is still built with Python 2 because it needs Mercurial (which is currently Python 2).

debian: update debhelper compat Level before 9 are deprecated says debhelper 12. Upgrade to 10, which is available in Debian oldstable and old Ubuntu releases to be conservative (and since we probably don't need more).

debian: make it possible to by-pass doc build

debian: define extend-diff-ignore for dpkg-source This makes it possible to build the source package from a hg checkout (and a sdist) by having dpkg-source ignore files not distributed in sdist.

debian: use pybuild and dh_sphinxdoc Use pybuild because it's kind of standard nowadays. Use sphinxdoc debhelper to have a Depends: libjs-sphinxdoc and common files (JS/CSS) symlinked.

debian: mark 9.2.1-1 as released

obsexchange: delete dead code for old exchange protocol Dead since 79a926b557f1 (compat: drop support for older exchange protocol, 2017-10-25).

branching: merge with stable

branching: merge the stable branch with the mercurial-5.1 branch This revert all the test changes to match the expected output for mercurial 5.1.

tests: add the new "-r" bit to the usual merge hint CORE-TEST-OUTPUT-UPDATE: 8197b395710e

test-output: reinstalled test change for 5.2 This effectively reinstall: * 8345b852cd4f * e804d5a7c193 * 62b60fc1983d * 2d85de79ead8 This mean backing out the following backout changeset: * 2eab4363fc62 * c1cc5b1f813a * 015cdd3fa469 * 3ba1c9dbf462

test-compat: opening test compatibility branch for mercurial 5.1 5.2 has a release candidate.

py3: encode/decode with going through ui in rst directive mercurial's ui wants bytes, doctutils wants text.

py3: fix setup.py --version Before: $ python3 setup.py --version b'9.3.0.dev' After: $ python3 setup.py --version 9.3.0.dev

rewind: preserve date

metaedit: don't change commit date by default (issue5994) We deliberately unset devel.default-date to allow hg commands to see the real system date, that demonstrates that metaedit doesn't fall back to system date for any of these commands.

tests: simply use testlib/common.sh in test-metaedit.t

tests: turn glog shell function into an alias in test-metaedit.t Mostly because we already have an alias and because creating an alias is the usual way to customize log -G in tests.

tests: clean up test-metaedit.t Mostly just artifacts from the times when it was one big test-evolve.t. All completely unused nowadays.

help: categorizing evolve and topic commands This makes them show up under the right categories in 'hg help'. Differential Revision: https://phab.mercurial-scm.org/D6999

tests: add the new "-r" bit to the usual merge hint CORE-TEST-OUTPUT-UPDATE: 8197b395710e

branching: merge back into default

pick: don't create any successors when there were no changes (issue6093) For example, when a changeset that user is trying to pick was already grafted or rebased (and no obsmarkers were created), pick will naturally say that there's no changes to commit and will not create the new commit. In that case it should not mark any changeset as a successor of the picked changeset, because the actual file changes were already there before picking. Technically, pick could (and did before this patch) mark the wdir commit as a successor of picked changeset (may be useful in the most trivial cases when e.g. wdir is a graft of the required change), but there are cases when the picked change is introduced in an ancestor of wdir, or even in multiple ancestors piece by piece. Pick definitely shouldn't be trying to guess something this complicated, so with this patch it always marks the picked change as pruned if it gave no changes to commit.

packaging: mark as development version This avoid confusion.

Added tag 9.2.1 for changeset d989bab720e5

test-compat: merge mercurial-4.6 into mercurial-4.5

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 stable into mercurial-5.0

packaging: prepare version 9.2.1

packaging: update changelog

manifest: exclude the gitlab file (oops)

branching: merge with stable

py3: return unicode from setup.py's get_version py3 expects the version number for eggs to be unicode, so despite 9.2.0 being the first evolve release with beta py3 support it isn't currently pip installable since it dies with: File "*/setuptools/command/egg_info.py", line *, in tagged_version return safe_version(version + self.vtags) TypeError: can't concat str to bytes The `setup.py install` test added in a previous commit should cover this since it died in a different but similar way before.

py3: remove smartquotes from README They break running setup.py with py3 but only in tests, because in the test environment `locale.getpreferredencoding()` is ANSI_X3.4-1968, so the file is `open()`ed with that as the encoding. An alternative for fixing this would be to make the `open()` call with `encoding='utf-8'` under py3, which would be safe against any future non-ascii in README.

setup: make runnable from other dirs Currently it fails when run in that way for two reasons: - the description is loaded from the README file but with a path relative to the working directory - module references have the same issue, which is fixed with package_dir.

setup: add a test for running setup.py `python3 -m pip install hg-evolve==9.2.0` currently fails, despite 9.2.0 being the first beta py3 release, because of unicode issues with the version number. `setup.py install` is a proxy test for pip install since it also suffers from similar, however the test currently fails due to running setup.py from a different directory to the one it is in. Also add a test for `hg version -v` with evolve enabled since a naive first solution for the pip issue was to just change `__version__` to a `u''` string, but that busts version display in `version -v`, which isn't currently picked up by tests.

gitlab-ci: be bold and try to run the full test in them This is a basic version that should work reasonably well. The setup is a bit costly and the cleanup is not guaranteed yet.

gitlab-ci: basic test with flake8 Let us step into the future.

branching: merge back with stable version 9.2.0 have been released

branching: preserve test change on stable CORE-TEST-OUTPUT-UPDATE: cdf0e9523de1 CORE-TEST-OUTPUT-UPDATE: d7304434390f CORE-TEST-OUTPUT-UPDATE: 181ee2118a96 CORE-TEST-OUTPUT-UPDATE: 181ee2118a96

packing: mark as development version This avoid confusion when installing from source.

Added tag 9.2.0 for changeset 40795751be1c

test-compat: merge mercurial-4.6 into mercurial-4.5

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 stable into mercurial-5.0

changelog: update for coming 9.2.0

test: backed out test change applied in 2d85de79ead8

test: backed out test change applied in 62b60fc1983d

test: backed out test change applied in e804d5a7c193

test: backed out test change applied in 8345b852cd4f

branching: merge with stable

stable-doc: note about not storing smaller range

stable-doc: small typo

stable-doc: add more advanced examples

stable-doc: add multiples example for the simple cases

test-output: oops CORE-TEST-OUTPUT-UPDATE: 181ee2118a96

test: update part count in tests An off by one error got fixed upstream. CORE-TEST-OUTPUT-UPDATE: 181ee2118a96

evolve: check that relocating makes sense in _solvedivergent() (issue5958)

tests: demonstrate content divergence causing "relocate node on top of itself"

cmdstate: introduce a "saver" contextmanager and use it in place of save() Previously, the state was only saved in some paths out of these functions. This can be problematic, if the user ctrl-c's (or `kill -9`'s) the process, or we exit out of `relocate` for anything besides the "expected" reason, we won't record that we were in the middle of an evolve. One of our users has discovered that this leaves hg in a weird state; the user did something like this: ``` $ hg evolve <something goes wrong with the merge tool, hits ctrl-c> <deals with the merge conflicts> $ hg evolve --continue abort: no interrupted evolve to continue $ hg evolve abort: uncommitted changes # Note: commands.status.verbose=True is set. $ hg status M foo # The repository is in an unfinished *update* state. # No unresolved merge conflicts # To continue: hg update ``` The user did an `hg update`, but it didn't actually do anything besides take it out of the unfinished update state (the files were still dirty in the working directory).

cmdstate: avoid setting a default argument to a mutable object (`{}`) If there's ever more than one cmdstate for the lifetime of the process, this can cause surprising behavior where the later cmdstates pick up options from the earlier ones. I've not seen any evidence this is actually causing any issues, but it's subtle enough that it should probably be fixed to help save significant debugging time later.

cmdstate: switch to new-style classes This isn't a huge issue and isn't necessary on Python3, but we're still Python2 compatible and we should be consistent (and strongly prefer new-style classes); this is the only class that is not deriving from object that I could find.

branching: merge stable into default

docs: add example for the `prune` command Giving at least one example helps clear up the syntax, especially since most options will follow a very similar pattern.

tests: add some more actions to test-issue-6028, describe what's being done This patch does two things: it explains what's happening in the test file so it's easier to understand, and also it adds checks that make sure something like swapping merge parents is not accidentally breaking anything. The primary reason to touch this test file was that it was broken by a change in core's merge.graft() that erroneously swapped merge parents. Since only evolve uses merge.graft() for merge commits and there aren't any tests in core for it, let's test it here. Plus, this test case is pretty simple and these additional checks don't make it too complicated.

topic: fix some API to make it more robust The new code can deal with changes in upstream mercurial.

rewriteutil: move allowdivergence check to foldcheck() foldcheck() calls precheck() first and then performs more fold-specific checks.

test: change to output around transaction timing CORE-TEST-OUTPUT-UPDATE: d7304434390f

branching: merge with stable There have been multiple fix on stable.

stablerange: add some data from field testing This show that in number of stable range to track stay reasonable.

stablerange: add a debugstablerangecache command The command display get some data about stable range in a repository.

stablerange: add some documentation about the general concept

stablesort: add some field data about stable sort cache

stablesort: introduce a small debugstablesortcache command This give a small insight on the cache content.

stablesort: add some documentation for stablesort This should help people to understand what is going on here.

fold: check allowdivergence before folding obsolete changesets (issue5817)

tests: demonstrate that fold doesn't care about allowdivergence config option

stack: remove unnecessary copying of rdependencies rdependencies is not modified in any way in this method, so no need to copy it.

stack: make a deep copy of `dependencies` before modifying its items The algorithm later on in this method uses .remove() to remove individual elements from items in dependencies, which before this patch modified the cached property contents. So for further use that dictionary was in the form of {1: set([])}, i.e. all sets were empty. This deep copy block could be way simpler, but the problem is that sometimes we get lists of _succs() from evolvebits.builddependencies(). Note: this happens only in topic's stack version of builddependencies() and it looks like a suboptimal way to handle multiple successors (see evolve's counterpart function). stack.builddependencies method is removed, it has served its purpose (see the previous patch).