gawk pb (IGNORECASE)

[Christophe Bisière]

Hi List,

With:

Linux tiwa.vgam 2.4.20-gentoo-r5 #2 sam mai 17 16:05:53 CEST 2003 i686 Pentium 
III (Coppermine) GenuineIntel GNU/Linux
gcc (GCC) 3.2.3 20030422 (Gentoo Linux 1.4 3.2.3-r1, propolice)
GNU Awk 3.1.2

running

echo "a" | ./gawk -v IGNORECASE=1 "/[[:alnum:]]/{}"

gives

gawk: cmd. line:1: (FILENAME=- FNR=1) fatal error: internal error

Notes:
- input does not matter
- the key point is the reference to the regexp while IGNORECASE=1
(the problem still appears when the regexp is in a sub, gsub, gensub, etc)

I tracked down the problem to the macro BUILD_CHARCLASS_LOOP in 
regcomp.c:3249.

For i > 127, trans[i] contains negative numbers. The call to bitset_set (line 
3255 in regcomp.c) thus segfaults.

As a workaround, changing (line 3254)
int ch = trans ? trans[i] : i;
to
unsigned char ch = trans ? trans[i] : i;
solves the problem (at least for me).

Cheers, and many thanks for your work,

Christophe

-- 
Christophe Bisière
University of Toulouse, France