Front page | perl.perl5.porters |
Postings from February 2002
Hugo Van Der Sanden <hv@crypt0.demon.co.uk> writes: >Slaven Rezic <eserte@vran.herceg.de> wrote: >:I'm not sure whether this is a Perl/Tk or perl problem: >[...] >: >:In perl's Perl_regexec_flags function, a core will be dumped at line >:1427: >: bool do_utf8 = DO_UTF8(sv); > >That's a problem with the regexp engine. And Tk ;-) > >:The problem is fixed if I change the line to: >: >: bool do_utf8 = sv && DO_UTF8(sv); > >That's fine as a starting point, There are other references to sv further down - some guarded, some not (in my quick scan). >but I suspect we need the ability to >match on an SV-less UTF8 string as well. I have no real problem requiring an SV for this. I just patched Tk to construct one. What mislead me was that SV * parameter used to be called 'screamer' and was something to do with 'study'. I never understood that and did not known that SV * _was_ the string. When Unicode-aware Tk gets out it will only ever use this on UTF-8 strings. >Perhaps we should add a flag >for the purpose? Then the above could be: > bool do_utf8 = sv ? DO_UTF8(sv) : flags & REXEC_UTF8; > >Note also that this would require some fixup for the debugging (which >assumes an sv if utf8) and for screaminstr() calls (which assume an sv >if REXEC_SCREAM). > >Hugo -- Nick Ing-Simmons http://www.ni-s.u-net.com/Thread Previous