Re: Why do we test a build using g++?

On Tue, 30 Jun 2026 12:50:08 +0100
"Paul \"LeoNerd\" Evans" <[email protected]> wrote:

> I don't know if it makes sense to scatter lots of GCC_DIAG_IGNORE
> calls around every callsite, or instead to arrange for just one to be
> placed somewhere toplevel in a .h file. Though that might be
> considered too rude. Hrmm.
> 
> Thoughts continue...

Thoughts indeed are continuing on this github issue:

  https://github.com/Perl/perl5/issues/24553

I opened that as an issue so that it doesn't need to hold up any of the
*other* issues or PRs, if that's the only remaining problem in them.
I.e. it'd be nice if we could merge any of the otherwise-valid C code,
if the only outstanding problem is that they're not warning-free when
treated as if it was C++. We can handle that as a separate problem
since it already exists - and has existed since at least 2021.

The offending line in Socket.xs was effectively introduced by this
commit:

commit b1846e36c303aefcfd6b0560936088badcbab8e0
Author: Richard Leach <[email protected]>
Date:   Mon Jan 18 02:39:46 2021 +0000

    Upgrade Socket from 2.030 to 2.031

If nobody has noticed (or cared) in the intervening 5-and-a-half years,
I don't think it should be treated as urgent and blocking new features
in current PRs now.

-- 
Paul "LeoNerd" Evans

[email protected]
http://www.leonerd.org.uk/  |  https://metacpan.org/author/PEVANS