gh-84559: improve What's New entry for multiprocessing start method changes

[picnixz]

This is a proposal for an improved changelog. I am tempted to add it at the very top of the file in the "Summary -- release highlights" but I don't know if @hugovk is ok or not with it.

• Issue: multiprocessing's default posix start method of 'fork' is broken: change to `'forkserver' || 'spawn' #84559

---

📚 Documentation preview 📚: https://cpython-previews--128173.org.readthedocs.build/

[hugovk]

I am tempted to add it at the very top of the file in the "Summary -- release highlights" but I don't know if @hugovk is ok or not with it.

Yeah, this sounds like a big change we want people to know about, let's add a summary up there and link down to multiprocessing part.

I expect we'll expand the summary closer to the release, compare 3.13:

https://docs.python.org/3/whatsnew/3.13.html#summary-release-highlights

cc @AA-Turner

[jakirkham]

Just a note, on macOS the default is spawn (as it has been for a while). However the docs also note that POSIX uses forkserver by default. Though macOS also is a POSIX platform. Should this be disambiguated somehow?

[picnixz]

Though macOS also is a POSIX platform. Should this be disambiguated somehow

Ah yes, sorry. it should have been "On POSIX platforms, except macOS, [...]" or "On platforms other than macOS and Windows". Which one would be the clearer?

[gpshead]

I'm glad we have a full section in the multiprocessing docs to link to (hooray past us!). But to help people from a What's New standpoint given the issues people are raising about this being confusing, I think we should pick a couple different short example error messages users might see coming out of an application that needs modification to work with the new start method when using multiprocessing or concurrent.futures.

[picnixz]

I think we should pick a couple different short example error messages users might see coming out of an application that needs modification to work with the new start method when using multiprocessing or concurrent.futures.

What about improving the section in multiprocessing about the differences between fork and spawn/forkserver by adding some "common issues" examples (and we can add a link to that section in the What's New entries) and/or possible examples of how to upgrade from 'fork' to 'spawn/forkserver'?

[gpshead]

I think we should pick a couple different short example error messages users might see coming out of an application that needs modification to work with the new start method when using multiprocessing or concurrent.futures.

What about improving the section in multiprocessing about the differences between fork and spawn/forkserver by adding some "common issues" examples (and we can add a link to that section in the What's New entries) and/or possible examples of how to upgrade from 'fork' to 'spawn/forkserver'?

Even just a sentence in What's New saying something along the lines of "if you are seeing name errors of pickling errors coming out of multiprocessing or concurrent.futures when you upgrade to 3.14, read ...link-to-section..." would help people understand. Expanding further upon that kind of thing in the multiprocessing docs as common issues would be good.

[picnixz]

Even just a sentence in What's New saying something along the lines of "if you are seeing name errors of pickling errors coming out of multiprocessing or concurrent.futures when you upgrade to 3.14, read ...link-to-section..."

I'll have a look at what to link tomorrow and if I have the motivation, I'll do something for the mp docs.

[ngoldbaum]

Small ping that I ended up here after a discussion about this behavior and getting this docs improvement merged would help a lot. No rush of course but just don’t want this forgotten.

[picnixz]

Oh I forgot... I will try to make it by the end of the weekend. Let's make it as the next task. Thanks for the ping !

[bedevere-app]

GH-137156 is a backport of this pull request to the 3.13 branch.