[regex] BZ #522 Enable UTF-8 and ASCII optimizations outside glibc, take 2

[Paolo Bonzini]

This is a simpler version of the patch I submitted on October 27.  It 
avoids reimplementing locale_charset, because the only charset name it 
has to check against is UTF-8: the four possibilities UTF-8, UTF8, utf-8 
and utf8 should cover most OSes, and were anyway the same that the 
bigger patch covered.

This patch avoids checking against a list of known ASCII-superset 
character sets, because I think it is enough to check that "btowc (x) == 
(wchar_t) x" for 0<=x<=127.  We already have a loop executing btowc 256 
times, so the cost of this additional, more robust check is small.

As I told Bruno in private mail, I'm quite reluctant to make sed 
slower/faster depending on the presence of a file other than /bin/sed, 
so using localcharset.c is not my favorite option.  Bruno, are you ok 
with this (or do you dislike it less)?

Paolo
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: 01-recognize-utf8-and-ascii-outside-glibc-take2.patch
URL: <http://sourceware.org/pipermail/libc-alpha/attachments/20041108/8b20b40b/attachment.ksh>