On Mon, 26 Nov 2007, I wrote:
> Having said that, I notice there are still some differences between PCRE
> and Perl (though Perl also doesn't match 0654 as Arabic), and I will do
> some more double checking just in case there is a problem in PCRE.
It turns out that there was a bug in the generation of PCRE's tables
from the Unicode data. It did put some characters (a relatively small
number) into the wrong scripts. I have fixed the bug and committed the
changes to the svn repo. This might change some of the effects you
reported, but not all of them were wrong. However, many thanks for
bringing this issue up, as it has caused me to fix a nasty bug. All the
tests I have done with the patched code now agree with Perl.
I am hoping to get a new release out before the end of the year.
