gh-124083: Skip test_strsignal on NetBSD due to TypeError

[furkanonder]

• Issue: test_strsignal fails on NetBSD with TypeError #124083

[Zheaoli]

I think we need to link this PR to the NetBSD community. We don't know whether this issue is caused by the NetBSD in a specific version or this is a wide exist bug.

I think we need to hold on this PR before we confirm the bug scope

[picnixz]

If strsignal is not supported on this platform, then the configure check must have detected it. However, it appears that your C program works correctly (namely that strsignal is present), so HAVE_STRSIGNAL should be defined. This also means that the following somehow returns None.

    errno = 0;
    res = strsignal(signalnum);

    if (errno || res == NULL || strstr(res, "Unknown signal") != NULL)
        Py_RETURN_NONE;

Either errno is actually set differently or res is NULL. But this seems quite weird that res is NULL since the C program works. So errno is somewhat modified.

If HAVE_STRSIGNAL is not specified, then the switch we are using is wrong. We should also check that the constants being check are actually the correct ones. I don't have access to a NetBSD machine but something more delicate could be hidden.

[Zheaoli]

If HAVE_STRSIGNAL is not specified, then the switch we are using is wrong. We should also check that the constants being check are actually the correct ones. I don't have access to a NetBSD machine but something more delicate could be hidden.

I'm working on the root cause. This is a little bit wired

[vstinner]

@Zheaoli: There are more details in the issue: #124083 (comment) signal.strsignal() returns None for most signals.

[vstinner]

LGTM

[miss-islington-app]

Thanks @furkanonder for the PR, and @vstinner for merging it 🌮🎉.. I'm working now to backport this PR to: 3.12, 3.13.
🐍🍒⛏🤖

[bedevere-app]

GH-124223 is a backport of this pull request to the 3.13 branch.

[bedevere-app]

GH-124224 is a backport of this pull request to the 3.12 branch.

[vstinner]

Merged, thanks for the fix @furkanonder.

[furkanonder]

Merged, thanks for the fix @furkanonder.

You are welcome!

[picnixz]

I'm still convinced that the fact it returns None is something that should be investigated. But I can live with this fix for now

[vstinner]

I'm still convinced that the fact it returns None is something that should be investigated

strsignal() is basically not implemented on NetBSD: #124083 (comment)

What do you want to investigate?

[picnixz]

Then why does it work with the C example on the issue? The function seems present and it returns a non-NULL result! (or am I misunderstanding something?)

[vstinner]

Then why does it work with the C example on the issue?

If fails with errno 2.

[picnixz]

Oh! I thought the result would be NULL in this case! My bad. Ok so nothing to investigate. Sorry for the noise :')

[vstinner]

No need to be sorry, strsignal() on NetBSD has a strange behavior (return a string and report an error). IMO it's easier to skip the test, there is no need to spend too much time on it until NetBSD is fixed.