GH-93316: socketmodule: remove constraint on queue size for macOS and FreeBSD.

[devnexen]

For those, it does make senseso the queue is set to SOMAXCONN (128 on
macOS desktop maybe more on the server flavor ?).

[gvanrossum]

The code looks fine, the question is whether we should even keep the < 0 check for other platforms.

It would also be nice to have a test.

[ronaldoussoren]

The Single Unix Specification mentions that a negative backlog should be interpreted as 0: https://pubs.opengroup.org/onlinepubs/9699919799/functions/listen.html

That makes the current check unnecessary and save to remove (and also means that the stated behaviour on macOS and FreeBSD is non-compliant).

@devnexen : Do you have a reference that documents this behaviour of those to platforms? The manpage for listen doesn't mention this.

UPDATE: My earlier comment about non-compliance is incorrect, the spec says that < 0 should be interpreted as == 0, and == 0 may update the backlog to some implementation defined minimum value. That's what's happening here.

[devnexen]

The Single Unix Specification mentions that a negative backlog should be interpreted as 0: https://pubs.opengroup.org/onlinepubs/9699919799/functions/listen.html

That makes the current check unnecessary and save to remove (and also means that the stated behaviour on macOS and FreeBSD is non-compliant).

@devnexen : Do you have a reference that documents this behaviour of those to platforms? The manpage for listen doesn't mention this.

macOs mentions subtly in its manpage (128 max but that's macOs desktop) however confirmed in the libc https://github.com/apple/darwin-xnu/blob/2ff845c2e033bd0ff64b5b6aa6063a1f8f65aa32/bsd/kern/uipc_socket.c#L1061
FreeBSD mentioned more clearly it s ok to use a negative backlog.

[devnexen]

The code looks fine, the question is whether we should even keep the < 0 check for other platforms.

It would also be nice to have a test.

There is an existing test testing negative backlog.

[ronaldoussoren]

I don't think a change is needed here, at least not for macOS. The code you linked to sets the backlog to SOMAXCONN when the value is <= 0. Python's listen binding doesn't change that (we clamp negative values to 0, but that's still setting the actual backlog to SOMAXCONN).

Just dropping the whole block of code would simplify the code base a little, but I'm not sure if it is worth the trouble.

[devnexen]

I don't think a change is needed here, at least not for macOS. The code you linked to sets the backlog to SOMAXCONN when the value is <= 0. Python's listen binding doesn't change that (we clamp negative values to 0, but that's still setting the actual backlog to SOMAXCONN).

Just dropping the whole block of code would simplify the code base a little, but I'm not sure if it is worth the trouble.

So the BSD conforms and I did the change and triggered the tests on linux.

[ronaldoussoren]

Could you add tests that explicitly use listen(0) and listen(-1) and create a connection afterwards?

On first glance we don't have such tests as the moment and adding such a test would quickly show if one of the supported platforms misbehaves with a negative backlog.

[bedevere-bot]

A Python core developer has requested some changes be made to your pull request before we can consider merging it. If you could please address their requests along with any other requests in other reviews from core developers that would be appreciated.

Once you have made the requested changes, please leave a comment on this pull request containing the phrase I have made the requested changes; please review again. I will then notify any core developers who have left a review that you're ready for them to take another look at this pull request.