bug#16919: [PATCH] fix mismatch between dfa and regex for treatment of titlecase

[Paul Eggert]

Norihiro Tanaka wrote:

> I found difference between dfa and regex (glibc) treatment of titlecase.

Thanks for bringing this up, but I'm afraid that it appears that regex 
is buggy in this area.  The regex code does the match by converting 
pattern and text to uppercase, and then trying a match with uppercase. 
But this is incorrect for an example like the following, which uses 
'\(\)\1' to force using the regex code:

echo 'ς' | grep -i '\(\)\1σ'

This should output nothing, because terminal sigma is not the same as 
lowercase sigma even when case is ignored.  But since the uppercase 
counterpart of both characters is capital sigma, grep incorrectly 
outputs the terminal sigma.  The dfa code gets it right.

POSIX is muddy in this area, unfortunately, but I don't see any 
interpretation whereby ς and σ should match when case is ignored.