Open encukou opened 4 months ago
It might be better for the tool to live outside the cpython repo, like blurb.
It might be better for the tool to live outside the cpython repo, like blurb.
The tool for replacing version "next" in .rst files with the number makes sense to keep in repo to me (as your draft PR does for now) as it is small. Really not a lot more than could be done with some inscrutable sed
commands.
It's small, but there can still be bugs in it, and it'd be easier to deal with those if we don't need to backport them.
@hugovk You said elsewhere that you'd prefer the version-bumping tool to live outside CPython repo. Since it's so small, I'd add it (and its test) to https://github.com/python/release-tools directly. Would that work?
You're a RM, so you get to choose the bikeshed paint here :)
@hugovk You said elsewhere that you'd prefer the version-bumping tool to live outside CPython repo. Since it's so small, I'd add it (and its test) to python/release-tools directly. Would that work?
Yep, this sounds good 👍 🖌️
It's now in 3.13 & 3.12. I plan to wait for releases from those branches before offering this to RMs of the security-only ones. (This is obviously not a security fix, but as a feature designed to make backporting easier, it might get an exception.)
It worked well for 3.14.0a1, you can see the next
-> 3.14
changes in https://github.com/python/cpython/commit/8cdaca8b25e3354530d66cbccc622e0e914e8e8a 👍
Yup! Pre-releases and dot-0's do have slightly different behaviour though.
Feature or enhancement
Proposal:
In a PR to CPython, the
versionadded
,versionchanged
,versionremoved
,deprecated
,deprecated-removed
directives in documentation should currently be set to the upcoming release.This is inconvenient:
It would be good to treat this more like News entries, which live in a
next/
directory before a release, when the release manager bundles them up and assigns a version.Concrete proposal:
versionadded
& the others to expand the version argumentnext
to<version> (unreleased)
(e.g.3.14.0b0 (unreleased)
).next
with a given string (e.g.3.14
).(unreleased)
. The RM should be able to skip this test, in case of a false positive.Has this already been discussed elsewhere?
I have already discussed this feature proposal on Discourse
Links to previous discussion of this feature:
https://discuss.python.org/t/automating-versionadded-changed-markers-in-docs-to-expedite-prs/38423
Linked PRs
Related PRs
Discourse announcement