GH-132775: Fix argument parsing for _interpqueues.put()

[AA-Turner]

Bug introduced in #134440, cc @serhiy-storchaka@ericsnowcurrently. Also xref gh-137685.

A

• Issue: Improvements to test.support.interpreters.Interpreter #132775

[ericsnowcurrently]

Just be sure I understood, the problem is that the arg spec for _queues.put() has one too many items in it. Is that right?

I'm pretty sure the $p shouldn't be there. The "fallback" arg is an int flag ¹, so it should have an "i" in the spec. Thus, the spec for queuesmod_create() should not change and for queuesmod_put() it should be "O&O|ii:put".

FWIW, at the time, with the feature freeze at hand, I didn't worry about ironing out little API quirks (e.g. placeholder args) in the low-level module. It isn't exposed to users, right?

Footnotes

1. See resolve_fallback() in Modules/_interpreters_common.h. ↩

[ericsnowcurrently]

To be clear, I left room for other "fallback" options later, rather than it being a definite binary choice. Sorry that wasn't more clear. Then again, that decision was probably a premature accommodation for a possible future. I figure it doesn't matter much though.

[AA-Turner]

Just be sure I understood, the problem is that the arg spec for _queues.put() has one too many items in it. Is that right?

Yes, specifically that the format spec for PyArg_ParseTupleAndKeywords expected five args, but kwlist gave only four.

I'm pretty sure the $p shouldn't be there. The "fallback" arg is an int flag 1, so it should have an "i" in the spec. Thus, the spec for queuesmod_create() should not change and for queuesmod_put() it should be "O&O|ii:put".

I've updated the PR. I don't think we need the test anymore given this is the current behaviour -- we are just removing the errant $p. Please would you be able to review/approve the updated PR?

FWIW, at the time, with the feature freeze at hand, I didn't worry about ironing out little API quirks (e.g. placeholder args) in the low-level module. It isn't exposed to users, right?

Definitely the correct choice. _interpqueues is technically exposed to users by dint of being importable, but is entirely undocumented and of course a private module. This PR is just a correctness fix that Serhiy spotted, but it would nevertheless be good to include it in 3.14.0 candidate 3 / final.

A

[ericsnowcurrently]

LGTM

[miss-islington-app]

Thanks @AA-Turner for the PR 🌮🎉.. I'm working now to backport this PR to: 3.14.
🐍🍒⛏🤖

[bedevere-app]

GH-138034 is a backport of this pull request to the 3.14 branch.

[serhiy-storchaka]

And I think that it would be better to remove passing explicit -1 as the fallback argument to _queue.create().