evolve: evolve --divergent failed when merge had nothing to do (issue4950) Before this patch, if the divergence resolution lead to an empty commit, the transaction would be rolled back and the divergence not resolved. This patch allows empty commits in that case. It is legitimate to end up with an empty commit in that case, for example when divergent commits differ over their commit messages.

inhibit: use dirty revset trick to speedup the inhibit search space. See documentation for details. I'm a bit sad about this changesets.

inhibit: use 'repo.revs' instead of 'repo.set' This skip the creation of a changectx object.

_inhibitmarkers: turn the result from generator into a list A generator object is always True, we have to turn it into a list in order to be able to use it in the tests. Before this patch we were rewriting the 'obsinhibit' file all the time, this had a performance impact especially on repository where transaction are expensive (hooks, cache invalidation, etc). Writing this file less often have a minor effect on test where a stripped changeset does not get removed from the obsinhibit file is still inhibited when unbundled back. I think this is fine and I'm moving forward.

split: drop the experimental flag I've been using it without issue for some time. The only blocker in my mind was the lack of commit message preservation and this is now fixed.

split: preserve original changesets description Losing the description on split is to much an issue, we offer the original description to edition for every part of the split.

merge with stable

readme: updateing changelog We forgot a couple of recent improvement.

evolve: rewrite command help (evolve command only)

parents: avoid locking the repository during 'hg parents' The wrapping code was initially written for update and pull who need the lock anyway. We duplicated the logic in the parent case to remove the need for locking.

README: update instructions for running tests.

tests: remove the custom run-tests.py This test-runners is a copy from the Mercurial one and slowly drift out of sync. Use the Mercurial's core test runners instead.

make: clean test related target We removed the runner, so they are broken. We can reintroduce them later with proper data to fetch an official 'run-tests.py' runner.

test: use 'dummyssh' from the run-tests directory This is a step toward not shipping our own test runners and relying on the official Mercurial one.

test: use 'killdaemons.py' from the run-tests directory This is a step toward not shipping our own test runners and relying on the official Mercurial one.

merge back with stable

Added tag 5.2.1 for changeset c15d6168412f

prepare version 5.1.2

merge with default there is some 3.6 related fix and test change that we need on default. No other significant change happened since last release expect for split, still marked experimental. So we prepare for a bugfix release.

split: mark the command as experimental It has a bit too many rough edge for now.

tests: run test with 3.6 and apply test change Tested with 3.4, 3.5 and 3.6. Some minor test update had to be done.

prune: have dirstate collaborate with transaction With 3.6, dirstate update can work with transaction to ensure proper rollback. We do it if the mercurial version permit it.

evolve: have dirstate collaborate with transaction With 3.6, dirstate update can work with transaction to ensure proper rollback. We do it if the mercurial version permit it.

prune: reuse revset from repair instead of copy-pasting code

merge with stable

evolve: compatibility with coming 3.6 Some function in the import stack changed, we have to adapt.

evolve: be more complete about copying extra from old revisions For 'hg evolve' we want to copy all extra-level metadata except for branch. This keeps the experimental topics extension from getting obliterated by evolve runs. Many hashes in tests changed due to some extra data coming along for the ride in various places, but no functionality that was tested. It's plausible that this copies too much metadata - maybe we should be dropping rebase_src or convert_revision as well. Right now no tools do that, so I'm biasing towards preserving everything except branch when copying metadata for now.

evolve: ignore ui argument passed to help loader The API changed at Mercurial e0c572d4d112.

directaccess: inspect trees of len() > 3 Previously, when inspecting revset AST's we'd only traverse down the tree if it was length 3 ([op, left, right]). In some situations, like 'or' the tree node will be greater than length 3 ([op, first, second, ..., nth]). So we need to traverse all the parts of the node to catch all the symbols.

debugobsconvert: make sure obsstore is loaded before version comparison hg 44918682093f delays loading obsstore, so we have to access _all to get the correct store version.

evolve: use repo._bookmarks.recordchange instead of repo._bookmarks.write We want to get rid of the api repo._bookmarks.write and this patch removes its use in evolve.py. Before this patch, we were using repo._bookmarks.write to save bookmarks change immediately instead of repo._bookmarks.recordchange that write change when transaction ends.

evolve: indentation change for making next patch more legible In the next patch: "evolve: use repo._bookmarks.recordchange instead of repo._bookmarks.write" we need to add a transaction in the rewrite function. To do so adds an indentation level and makes the patch harder to review. This patch makes the indentation change so that the next patch is easier to review.

merge with stable

next/prev: properly lock repo before touching bookmarks

next/prev: properly lock repo before touching bookmarks

prune: ensure we prune bookmarks in a transaction

next: add support for --dry-run also adds support to prev

evolve: spelling uncommitted

evolve: annotate translatable strings with _()

evolve: do not capitalize description of next/previous

evolve: _() should not contain %

evolve: add progress for _obsrelsethashtree

evolve: add progress for the obsmarkers discovery parts

test-obsolete: clarify divergence detection output order

directaccesss: also look for symbol into list revrange is now much smarted and optimise multiple simple symbol (like hash) into an efficient '_list' call. So we need to look into these to find user provided input too.

merge with stable

inhibit: fix inhibit working with non-inhibit repos Inhibit was breaking when two repos were in memory, but one was not an inhibit repo (like when doing a local pull between two repos). The fix is to add inhibitenabled checks to all the places where inhibit does global wrapping of commands (every code path from the extsetup wrappers).

directaccess: don't try to partialmatch things that aren't hashes Trying to partialmatch identifiers like '.' turns out to be (a) pointless and (b) extremely slow. On a repo with a million commits, with directaccess enabled, hg log -r .^::. goes from 2.1 seconds to 0.5.