Mercurial 5.2 release as stable release for Python 3.

[Pulkit Goyal]

On Mon, Oct 14, 2019 at 8:08 PM Gregory Szorc <gregory.szorc at gmail.com> wrote:
>
> I support marking the code base as stable with Python 3 in the upcoming few weeks - at least for non-Windows.
>
> If we're serious about this, we all need to be running Mercurial with Python 3 locally and fixing bugs. I tried installing a Python 3 Mercurial a few days ago and I encountered enough tracebacks to cause me to revert. Those include https://bz.mercurial-scm.org/show_bug.cgi?id=6196 and an issue with evolve/obsmarkers that can be reproduced by `hg push` to hg-committed. I think we should land a patch to Makefile that changes https://www.mercurial-scm.org/repo/hg/file/649a9601b9e2/Makefile#l8 to `python3` to force the issue. We can revert that before tagging 5.2 if we want to do a separate release that is Python 3 primary (there was talk of doing a 2.7 5.2 then doing a 5.2.1 or a 5.3 a week or two later that is Python 3 native).
>
> We also still have a handful of test failures on Python 3. See https://ci.hg.gregoryszorc.com/. (CI is broken with Python 3.5 for some reason. I'll look into it.)
>
> Python 3 on Windows has a ton of test failures. ~120 I believe. I suspect a lot of them are the same underlying issues. I almost have my CI system working with Windows. But we'll certainly need a bit of effort on Windows before we can consider Python 3 stable there. At this time, Python 3 on Windows seems to be at risk because of the volume of issues.
>
> Other major blockers to Python 3 are packaging work. I'm the de facto maintainer of the Windows packages and will write those patches. But my priorities are standing up Windows CI because I think having visibility into the test failures is more important because what good is Python 3 packages if Mercurial isn't usable :) I could also help with non-Windows packaging if it is needed.
>
> I would also propose we reinstate the @ code freeze for this release so we can all focus on Python 3 and quality of life improvements for the release. I don't think we'll put out a quality Python 3 release if we're distracted by feature work on @.

I am big +1 on this one.