Front page | perl.perl5.porters |
Postings from December 2008
On Dec 01 2008, Chip Salzenberg wrote:
> On Mon, Dec 01, 2008 at 11:52:03PM -0500, Rick Delaney wrote:
> > diff -pruNb perl-5.8.x/scope.c perl-5.8.9-RC1/scope.c
> > --- perl-5.8.x/scope.c 2008-10-29 04:23:15.000000000 -0400
> > +++ perl-5.8.9-RC1/scope.c 2008-12-01 23:27:07.000000000 -0500
> > @@ -621,12 +621,17 @@ Perl_leave_scope(pTHX_ I32 base)
> > const int type = SSPOPINT;
> > switch (type) {
> > case SAVEt_ITEM: /* normal string */
> > + {
> > + bool was_tainted = 0;
> > value = (SV*)SSPOPPTR;
> > sv = (SV*)SSPOPPTR;
> > + was_tainted = SvTAINTED(value);
> > sv_replace(sv,value);
> > + was_tainted ? SvTAINTED_on(sv) : SvTAINTED_off(sv);
> > PL_localizing = 2;
> > SvSETMAGIC(sv);
> > PL_localizing = 0;
> > + }
> > break;
> > case SAVEt_SV: /* scalar reference */
> > value = (SV*)SSPOPPTR;
>
> Since SvSETMAGIC() could set or clear the taint flag, perhaps any manual
> manipulation of that flag go _after_ the SvSETMAGIC?
Good point, although in this case magic_settaint() will never even get
called because PL_localizing is 2 (see mg_set()). Not sure if we have
a regression test for this, though, or how to even write one.
--
Rick Delaney
rick@bort.ca
Thread Previous