grep: matching '\([A]\|[B]\)\{2\}' in different locales

[Harald van Dijk]

Hi,

When I try to grep '\([A]\|[B]\)\{2\}', and as input I pass AA / AB / BA / 
BB (yes, on separate lines), in the C locale, grep prints all four lines. 
However, in UTF-8 locales (specifically en_GB.UTF-8, but I got the same 
result with some others as well), it only prints the second and fourth line. 
If I change [B] to B, or [A] to A, or both, grep prints all four lines 
regardless of locale. If I remove \{2\} and try '\([A]\|[B]\)\([A]\|[B]\)' 
instead, grep again prints all four lines regardless of locale.

I've tested this with grep 2.5.1 on Linux, configured with nothing more than 
--prefix=$HOME/grep, compiled with gcc 4.0.1, on a system with glibc 2.3.5 
(20050421 snapshot). I also tried --with-included-regex just in case, but 
this didn't change anything.

Am I correct in assuming this is a bug?