socketmodule sock_listen accepts backlog negative arg on macOs and FreeBSD

[devnexen]

Feature or enhancement

For these systems, it is set implicitly SOMAXCONN.

Pitch

Allowing negative value for those systems since it s allowed and makes sense, recent Linux kernels versions can too but it can enabled in another time.

Linked PRs

• GH-93316: socketmodule: remove constraint on queue size for macOS and FreeBSD. #93315

[AA-Turner]

Hi @devnexen,

Per the issue template, please give a clear and concise description of your proposal -- what do you want to change in socket from the current status quo? We'll close this issue if we don't hear back from you.

It may be helpful to raise this on the Python discourse first.

A

[gvanrossum]

I'm guessing that the OP would like some change to these lines

    /* To avoid problems on systems that don't allow a negative backlog
     * (which doesn't make sense anyway) we force a minimum value of 0. */
    if (backlog < 0)
        backlog = 0;

in socketmodule.c, in sock_listen(), since the kernel on the mentioned OS'es supports this (I haven't verified that).

The OP posted https://discuss.python.org/t/gh-93316-socketmodule-allow-negative-backlog-for-macos-and-freebsd/16055 but did not provide extra details. (It seems they are asking for someone to help them develop a PR?)

I'm not sure how realistic such a PR is. We don't know exactly which OS'es support this, and you can get the same functionality by explicitly passing socket.SOMAXCONN. Unless the dynamic limit is sometimes different from the defined constant? I've heard of such things, but I don't know if this is the case here.

@devnexen could you link to (or quote from) online docs explaining the OS feature you are referring to?

[AlexWaygood]

I think OP already created a PR for this issue (#93315), but created it before the issue. Since we haven't got autolinking between issues and PRs up and running yet (post-GitHub migration), there was no way of knowing that from this issue alone.

[ronaldoussoren]

I've commented on the PR: The usefulness of the PR is limited, as macOS treats "<=0" the same as SOMAXCONN. This means the PR has no effect.

We could drop the clamping of backlog entirely if we can assume that all platforms are standards conforming (as the standard says that "< 0" is the same as "== 0".

[devnexen]

Thanks, so other BSD do it too, agree of dropping the double work on python's side.

[gvanrossum]

I’d be okay with just dropping the clamping.