On Thu, Jan 28, 2010 at 12:49:20PM -0500, Tom Lane wrote:
> Joe Conway <[email protected]> writes:
> > I pull directly from CVS, not git, but in any case my line 1117 is
> > subref = newRV_inc((SV*)GvCVu((GV*)sub_glob));
> > so it appears to be the same
>
> >> What perl version are you using?
> >> What compiler version are you using?
> > I'm on stock Fedora 12:
>
> I see the same on Fedora 11. The -E expansion of the line in question is
>
> subref = Perl_newRV(((PerlInterpreter
> *)pthread_getspecific((*Perl_Gthr_key_ptr(((void *)0))))),
> (SV*)((((GV*)sub_glob)->sv_u.svu_gp)->gp_cvgen ? ((void *)0) :
> (((GV*)sub_glob)->sv_u.svu_gp)->gp_cv));
>
> so it's evidently unhappy about the fact that GvCVu can return null,
> while Perl_newRV is declared __attribute__((nonnull(2))).
>
> It looks to me like this is probably a live bug not just compiler
> hypersensitivity.
Yes. (ISTR there have been cases where the notnull attribute was
misapplied to some perl functions, but that's not the case here.)
I think I missed this because the Xcode compiler on Snow Leopard is
fairly old (gcc 4.2.1).
Patch attached.
Tim.
diff --git a/src/pl/plperl/plperl.c b/src/pl/plperl/plperl.c
index 9277072..2dd3458 100644
*** a/src/pl/plperl/plperl.c
--- b/src/pl/plperl/plperl.c
*************** plperl_create_sub(plperl_proc_desc *prod
*** 1113,1120 ****
if (count == 1) {
GV *sub_glob = (GV*)POPs;
! if (sub_glob && SvTYPE(sub_glob) == SVt_PVGV)
! subref = newRV_inc((SV*)GvCVu((GV*)sub_glob));
}
PUTBACK;
--- 1113,1123 ----
if (count == 1) {
GV *sub_glob = (GV*)POPs;
! if (sub_glob && SvTYPE(sub_glob) == SVt_PVGV) {
! SV *sv = (SV*)GvCVu((GV*)sub_glob);
! if (sv)
! subref = newRV_inc(sv);
! }
}
PUTBACK;
--
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
