https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96122
--- Comment #5 from Paul Thomas <pault at gcc dot gnu.org> ---
(In reply to Paul Thomas from comment #4)
> (In reply to Thomas Koenig from comment #3)
> > (In reply to Andrew Benson from comment #2)
> > > (In reply to Dominique d'Humieres from comment #1)
> > > > Confirmed. For me it segfaults from GCC7 up to GCC11.
> > >
> > > Interesting. This started occurring for me when I updated from GCC10.1 to
> > > 11. But, the code I posted is a reduced test case that I created - so
> > > maybe
> > > the original (non-reduced-test-case) code wasn't triggering this
> > > previously
> > > for some reason.
> >
> > In Gcc 10.1, we did not call the finalizer for all required instances
> > (PR 94109). This has been fixed for gcc11 and the upcoming 10.2 release.
> >
> > It is likely that calling a (required?) finalizer again has led to
> > the segfault reappearing.
> >
> > (Our finalizer handling is a mess, this and PDTs are really the
> > only two really buggy fields for F2003 conformance).
>
> The segfault is caused by the following:
> {
> struct array01_vs parm.6;
>
> parm.6.span = 64;
> parm.6.dtype = {.elem_len=64, .rank=1, .type=5};
> parm.6.dim[0].lbound = 1;
> parm.6.dim[0].ubound = 2;
> parm.6.dim[0].stride = 1;
> parm.6.data = (void *) &ptr1->n[0];
> parm.6.offset = -1;
> __final_a_Vs (&parm.6);
> }
>
> whereas the template for __final_a_Vs is:
>
> integer(kind=4) __final_a_Vs (struct array15_vs & restrict array,
> integer(kind=8) byte_stride, logical(kind=1) fini_coarray)
>
> at a minimum, 'byte_stride' is required in the call
>
> I have a mostly complete patch, for finalizers in general, which includes
> this and so fixes the segfault.
>
> I hope to submit the current version of the patch today because we have
> started stage 4.
>
> Paul
By the way, in order to trigger some finalization calls, 'vsf' had to declared
'elemental'. Without my patch, it still segfaults in runtime, even with this
embellishment.
Apart from PR65347 and PR88735, I have now fixed all of the finalization bugs
on which PR37336 depends.
Thanks for your help and bug reports.
Paul
