On Tue, Jun 07, 2005 at 11:15:03AM +0200, Rafael Garcia-Suarez wrote:
> Yitzchak Scott-Thoennes wrote:
> >
> > How do you feel about this:
> >
> > --- perl/av.c.orig 2005-06-02 01:07:41.000000000 -0700
> > +++ perl/av.c 2005-06-06 02:05:04.853572800 -0700
> > @@ -927,8 +927,8 @@ Perl_av_exists(pTHX_ AV *av, I32 key)
> > return FALSE;
> > }
> >
> > - if (key <= AvFILLp(av) && AvARRAY(av)[key] != &PL_sv_undef
> > - && AvARRAY(av)[key])
> > + if (key <= AvFILLp(av) && AvARRAY(av)[key]
> > + && (AvARRAY(av)[key] != &PL_sv_undef || ! AvREAL(av)))
> > {
> > return TRUE;
> > }
> >
> > which should make exists true for any parameter actually passed, so
> > long as @_ isn't reified.
>
> I don't like it. It doesn't solve the whole problem but only reduces
> the set of cases where weird behaviour is observed, making this weird
> behaviour perhaps more obscure and less predictable.
I agree with the sentiment, but think it would be worth it (perhaps in
conjuction with a short warning in perlfunc/exists that exists on @_
or @DB::args may not work).
The only complete fix I see would be to also make av_reify replace all
PL_sv_undefs with new sv's, something like (untested):
--- perl/av.c.orig 2005-06-02 01:07:41.000000000 -0700
+++ perl/av.c 2005-06-07 16:25:15.358164800 -0700
@@ -40,6 +40,8 @@ Perl_av_reify(pTHX_ AV *av)
assert(sv);
if (sv != &PL_sv_undef)
(void)SvREFCNT_inc(sv);
+ else
+ AvARRAY(av)[key] = newSV(0);
}
key = AvARRAY(av) - AvALLOC(av);
while (key)
This doesn't seem like too awful a penalty to pay for making exists()
work consistently.
Unless there's something that can be done to pp_entersub (probably too
expensive) or pp_undef (don't know the repercussions of this).
