[BZ558] Part 1: more efficient lowering of {m,n}

[Paolo Bonzini]

> Does this DTRT with say ([abcdefgh]){1,5}\1, e.g. when matching
> abcdee or abb? If yes, it would be good to have that in the testsuite
> too.

Yes, it does.  It is simply a different way of lowering: if matching a . 
failed, it is not productive to try the epsilon transition and try to 
match another one.

PCRE does this to lower repetitions of subexpressions, by the way.

>>Patch at 
>>http://sources.redhat.com/bugzilla/attachment.cgi?id=289&action=view
> 
> Even if you put patches into bugzilla, please mail them as well to
> the mailing list.

Ok, here it is.

Paolo
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: bz558-eclosure.patch
URL: <http://sourceware.org/pipermail/libc-alpha/attachments/20041123/fa85a594/attachment.ksh>