Subject: 5.9 freeze, Windows, regression-fixes and Python 2 drop

[Raphaël Gomès]

Hi all,

As you know, the 5.9 freeze is coming up and should happen on the 18th 
of July.

This release will be a special one since, as of yesterday¹, all critical 
tests pass on our Windows Python 3 Heptapod CI. Tests related to 
extensions (like SVN support, `remotefilelog`, etc.) still have to be 
worked on and there is minor flakyness with a few tests², but the 
message is: I vote that this be the first release to support Python 3 
Windows officially and the last to support Python 2 altogether. This 
will give us enough time to iron out the details in time for the 6.0 
release to be the first Python 3 only release of Mercurial.

For the past few versions we haven't had much of a real freeze, which I 
think makes sense given the volume of contributions and the relatively 
good track-record of (lack of) regressions, but this time I want to 
propose that we observe a real freeze to dedicate at least some time to 
fixing the issues we're having and stabilizing the new stuff.

To name some things explicitly:
   - The parent ordering issue³ which is almost 50% why I'm advocating 
for the freeze⁴
 Â  - Windows Python 3 support and improving CI coverage
 Â  - `chg` now almost consistently times out some tests
 Â  - `test-chg.t` is super flaky because of `sleep` calls
 Â  - Bugs that have began building a settlement in bugzilla

I, for one, am happy that the end of the Py2 era is finally visible at 
the end of the tunnel and would like to thank all involved in the recent 
Windows endeavor.

Lastly, the Heptapod Windows CI Octobus is providing for this transition 
will cost about 400€ per month. We're currently running on free credit 
we have thanks to a partnership with OVH given for use in the Heptapod 
project, but this will run out one day and we very much encourage those 
who can help to reach out. Octobus is a 5-people European team already 
funding the development of Heptapod out of pocket, and there is only so 
much that can be reasonable for us to shoulder in the long-run.

What are your thoughts?
Raphaël

[1] https://foss.heptapod.net/octobus/mercurial-devel/-/jobs/222914
[2] https://foss.heptapod.net/octobus/mercurial-devel/-/issues/12
[3] https://bz.mercurial-scm.org/show_bug.cgi?id=6538
[4] Larger users, including Mozilla, are affected by the bug: 
https://bugzilla.mozilla.org/show_bug.cgi?id=1720302