Okay, I did a bit of comparison. The machine the binary was *compiled* on seemed to matter, not the one it was *run* on. A bit of extra printing revealed the obvious culprit:
[539]$ ./a.out sizeof regex_t = 28 array[1] confirmed all-zero J'accuse! buf[0] = 0x18 Bug! array[1] has been overwritten! [524]$ ./a.out sizeof regex_t = 32 array[1] confirmed all-zero array[1] confirmed all-zero It's obviously a header file issue. I will go investigate. The /usr/include/regex.h files are identical between the machines, it must be something else... Aha! Where did /usr/local/include/sys/types.h come from? It's been there since 2007 and never caused a problem until now... Anyway, the difference seems to boil down to the following gcc -E difference: (- = buggy, + = works) unsigned char *__buffer; unsigned long int __allocated; - unsigned long int __attribute__((__used__)); + unsigned long int __used; reg_syntax_t __syntax; Sorry to jump to conclusions and blame you. Thanks for trying to reproduce and putting me on the right track! -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
