On Fri, May 14, 2004 at 09:43:03AM +0200 Rafael Garcia-Suarez wrote:
> Tassilo von Parseval wrote:
> >
> > suppose I want to trigger external Perl subroutines but don't want to
> > pass the argument via the stack but rather by setting $_. Is the
> > following correct?
> >
> > void call_coderef (SV *coderef, char *path) {
> > SV *src;
> > dSP;
> >
> > SAVESPTR(DEFSV);
>
> or in other words SAVE_DEFSV
I read this in pp_grepstart:
/* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
SAVESPTR(DEFSV);
so I used it, too. Of course, I am not yet using pTHX_ which I probably
would have to in this case.
> > ENTER;
> > SAVETMPS;
> > PUSHMARK(SP);
> > DEFSV = sv_2mortal(newSVpvn(path, strlen(path)));
> > PUTBACK;
> > (void) call_sv(coderef, G_DISCARD);
> >
> > FREETMPS;
> > LEAVE;
> > }
>
> That looks correct to me.
Cool.
> A caveat, though: this will not work with a lexical $_ in perl 5.9.x.
> DEFSV is always $::_. With a lexical $_ (you can know whether there is
> one by looking it up in the pad) you have to replace SAVE_DEFSV by
> SAVESPTR(PAD_SVl(padoffset_of_dollarunderscore)). Grep bleadperl for
> details.
>
> Hmm, maybe this means that a _new_ macro should be added for XS
> writers, that get DEFSV right, by testing whether there's a lexical $_
> in scope. Opinions ?
Well, yes, please! Ideally, it would be backwards-compatible so that
people could already start using it now (unless of course they
deliberately want $::_).
Tassilo
--
$_=q#",}])!JAPH!qq(tsuJ[{@"tnirp}3..0}_$;//::niam/s~=)]3[))_$-3(rellac(=_$({
pam{rekcahbus})(rekcah{lrePbus})(lreP{rehtonabus})!JAPH!qq(rehtona{tsuJbus#;
$_=reverse,s+(?<=sub).+q#q!'"qq.\t$&."'!#+sexisexiixesixeseg;y~\n~~dddd;eval
