Front page | perl.perl5.porters |
Postings from December 2000
Nick Ing-Simmons <nik@tiuk.ti.com> writes:
>Nick Ing-Simmons <nik@tiuk.ti.com> writes:
>>B. When I fix A in Tk sources it core dumps in a manner which suggests
>> something has done heap-overrun.
>
>Debugging this I discovered a "feature" of sv_utf8_upgrade:
What a synchronicity of issues !
I posted this:
>
>void
>Perl_sv_utf8_upgrade(pTHX_ register SV *sv)
>{
> char *s, *t;
> bool hibit;
>
> if (!sv || !SvPOK(sv) || SvUTF8(sv))
> return;
>
> /* This function could be much more efficient if we had a FLAG in SVs
> * to signal if there are any hibit chars in the PV.
> */
> for (s = t = SvPVX(sv), hibit = FALSE; t < SvEND(sv) && !hibit; t++)
> if (*t & 0x80)
> hibit = TRUE;
>
> if (hibit) {
> STRLEN len = SvCUR(sv) + 1; /* Plus the \0 */
> SvPVX(sv) = (char*)bytes_to_utf8((U8*)s, &len);
> SvCUR(sv) = len - 1;
> SvLEN(sv) = len; /* No longer know the real size. */
> SvUTF8_on(sv);
> Safefree(s); /* No longer using what was there before. */
> }
>}
and
Nicholas Clark <nick@ccl4.org> writes:
>How come sv_2iv has
> if (SvREADONLY(sv) && SvFAKE(sv)) {
> sv_force_normal(sv);
> }
>
>but sv_2uv and sv_2nv don't have the same code in the similar position?
And the Tk bug is related to both!
For reasons that are not clear to me SPARC/gcc-2.95.2 is mis-compiling
that code:
hibit=262144 s=283410 t=283411 e=283416
SV = PVIV(0x393468) at 0x397e34
REFCNT = 1
FLAGS = (TEMP,POK,FAKE,READONLY,pPOK)
IV = 1941317403
PV = 0x283410 "-class"
CUR = 6
LEN = 0
Thus I am getting false +ve 'highbit' on the '-' character.
This in turn leads to the string being upgraded and then Safefree'ing
something which it shouldn't - hence SEGV later.
--
Nick Ing-Simmons <nik@tiuk.ti.com>
Via, but not speaking for: Texas Instruments Ltd.