bpo-41808: Add What's New 3.9 entry missing from master

[terryjreedy]

Entry was added by bpo-40939, #21012 and #21039.

https://bugs.python.org/issue41808

[terryjreedy]

Is addition correct, or should entry be deleted from 3.9 copy of same file.

[pablogsal]

Is addition correct, or should entry be deleted from 3.9 copy of same file.

Is correct, those functions are already gone in master and they are deprecated in 3.9

[bedevere-bot]

@terryjreedy: Please replace # with GH- in the commit message next time. Thanks!

[gvanrossum]

Please add the 3.9 back port label

[terryjreedy]

This is a forward port of what was added to 3.9 already.

[gvanrossum]

Unusual workflow.

[terryjreedy]

Our versioning of each "What's New x.y", combined with our branching of master at .b1*, requires we thereafter exactly repeat ourselves at least once to change what is a conceptually an unversioned document. This while the document is still in very much in development.

This workflow obviously violates "Don't repeat ourselves" (without necessity). I think putting What's New in a separate repository, as with the devguide, would be worse. Perhaps better, at least for PR writers, would be to have the the What's New directory only in master, by removing What's New in x.y from the x.y branch as soon as it is created. This would require the online 3.x doc page to pull What's New in 3.x from master instead of the 3.x branch. Maybe there would be other problems.

The chance of repetition error is reduced by auto backports. This PR fixes 1 of the 2, but only 2, inconsistencies git diff found in "What's New 3.9". Bpo-40939 had 17 merges to master and 3.9. Because most had to be different for master and 3.9, only 1 was a miss-islington backport. The What's New in 3.9 entry in question here was added and edited as part of PRs** that otherwise only applied to 3.9. It should instead have been added by a separate PR for master and then backported. I don't blame a relatively new contributor for missing this subtlety.

* Not too many versions ago, this branching was done much later, at .rc1 I believe (which require suspension of enhancement commits).

** 21039 in the commit message should have been #21036. My error. I should have clicked on the links to check for typos before merging.

[gvanrossum]

So IIUC the recommended workflow is to update "whatsnew x.y" in master and use the auto-backport machinery, and that's what usually happens, but occasionally that fails because the code part of a certain PR is just too different in the backport branch and not all developers/contributors are aware of the subtle workflow. I understand completely. Thanks for explaining!