regex.h problem

[Dana Lacoste]

There seems to be a bug in regex.h when it comes to handling
escaped single quotes (talk about obscure :)

specifically, the test program below fails on glibc 2.2.3
but passes on 2.1.2 without a problem.

I'm currently trying to see if I can get other glibc's tested
(FreeBSD 4.5's libc passes, but I know that's not useful :)

(There's sample code attached at the bottom of the message)

The specific situation : reading a string that occasionally
has a single quote in front of it, so trying to pattern match
on an optional single quote.

This is a bug because if you switch the ' character with
any other character (say, an X) it works fine.  It also works
if the ' character is not escaped, and if the ' character
is escaped inside a set of \( \) parentheses.

To summarize :
"^\\(\\'\\)?STUFF" passes
"^'?STUFF" passes
"^\\X?STUFF" passes
"^\\'?STUFF" fails

Using gcc 2.95.3, glibc 2.2.3

Dana Lacoste      - Linux Developer
Peregrine Systems -  Ottawa, Canada

-- begin sample/test code --

#include <stdio.h>
#include <regex.h>

main(int argc,char **argv)
{
 const char *txt = "^\\'?STUFF";
 regex_t reg;
 int err_code = regcomp(&reg,txt, REG_EXTENDED | REG_NOSUB);
 if(err_code != 0)
 {
    size_t errbuf_size = regerror(err_code,&reg,NULL,0);
    char *errbuf = new char[errbuf_size];
    regerror(err_code,&reg,errbuf,errbuf_size);
    printf("Error = (%d)%s (%s)\n",err_code,errbuf,txt);
 }
 else
 {
    size_t errbuf_size = regerror(err_code,&reg,NULL,0);
    char *errbuf = new char[errbuf_size];
    regerror(err_code,&reg,errbuf,errbuf_size);
    printf("Ok = (%s)\n",txt);
 }
}