Re: perl v5.42 segfault

[Thomas Klausner <wiz%gatalith.at@localhost>]

On Wed, Oct 15, 2025 at 04:54:18PM +0100, James Cook wrote:
> James Cook <falsifian%falsifian.org@localhost> wrote:
> > mlelstv%serpens.de@localhost (Michael van Elst) wrote:
> > > falsifian%falsifian.org@localhost (James Cook) writes:
> > > 
> > > >	perl -E 'for my($x,$y)($a->foo){}'
> > > 
> > > 
> > > This looks like a variant of:
> > > 
> > >    https://github.com/perl/perl5/issues/23405
> > > 
> > > 
> > > The issue mentions a fix, but unclear if that can be backported.
> > 
> > Oops, I missed your email. Thanks for finding it.
> > 
> > I am surprised this made it to a released version of perl. It seems quite
> > easy to trigger: you just need to use the "indexed" function (lexically
> > imported) in the most natural way imaginable.
> > 
> > Maybe I'll ask #perl on libera.chat for advice about backporting the fix.
> 
> Update: cherry-picking commit 96673a4bb36 on top of perl's v5.42.0 tag
> fixes the problem. I can confirm that perl's "make test" passes. People
> in #perl on libera.chat seem to think this is a reasonable approach.
> 
> I have not tried applying the fix in pkgsrc, but hopefully it should be
> as simple as adding a patch from commit 96673a4bb36.

I've tried applying just the single patch, but the perl built this way still says:

# ./work/.destdir/usr/pkg/bin/perl -E 'for my($x,$y)($a->foo){}'
Memory fault (core dumped)

 Thomas