view .gitlab/issue_templates/new-version.md @ 6224:17ffdea0edbb stable

evolve: look for split successors of the correct ancestor (issue6648)

Consider two changesets, 1 and 2. 1 is split into two new changesets and 2 is
pruned. If we stand on 2 and call hg evolve, _singlesuccessor() will traverse
ancestors of wdp in search of a changeset with successors to update to (it will
be 1, which was split). In case of a split, select_split_successor() gets
control. The issue is this function didn't traverse ancestors, and instead
tried to look up successors of the original changeset (i.e. 2 in our case,
which was pruned).

We can make select_split_successor() aware of _singlesuccessor() logic by using
the changeset that actually has successors without traversing ancestors again.
It's done by storing that changeset in MultipleSuccessorsError exception.

This is the actual check list for releasing evolve version X.Y.Z

More details in the [README.rst file](README.rst#L210).

Preparation

* [ ] merges with compatibility branches
* [ ] compatibility tests are happy
* [ ] update `testedwith` field
* [ ] make sure CHANGELOG is up-to-date
* [ ] [prepare an announcement for the list/website](https://mypads.framapad.org/p/evolution-announce-iswcp7qx)
* [ ] sanity check install and sdist targets of setup.py

Releasing

* [ ] add the current date to CHANGELOG
* [ ] update the `__version__` field of all relevant extensions
* [ ] create a new Debian entry
* [ ] tag the version
* [ ] move the `@` bookmark to the new tag
* [ ] push/publish the tag to the main repository
* [ ] upload the tarball to PyPI
* [ ] add `.dev` to the `__version__` field
* [ ] merge stable into default
* [ ] push the result to https://www.mercurial-scm.org/repo/evolve/
* [ ] send the announcement by email to [e-mail]
* [ ] send the announcement by email to [e-mail]
* [ ] publish the announcement as a blog on https://octobus.net/blog/
* [ ] update #hg-evolve topic
* [ ] tweet about it