This is the regex functionality from glibc 2.5 extracted into a
separate library, for Win32.

I call the DLL libgnurx-0.dll which hopefully should be unique. At
least it isn't "regex.dll" which has been used by the
gnuwin32.sourceforge.net site for *two* incompatible DLLs. (That mess,
and the mess with their build of Henry Spencer's regex library, was
what lead me to build my own GNU regex library. See the gnuwin32-users mailing list archives from December 2006.)

The "-0" is so that if at some point I build a release that isn't
binary compatible, I can then increment that and use a different name.

The import library for gcc is called libgnurx.dll.a, but I also
distribute a copy of it called libregex.a so that configure scripts
that look for -lregex will work.

Note that none of the wide-character and i18n functionality which is
built when this is part of glibc gets compiled. Thus things like
character classes most probably work only for single-byte codepages.

Compiling that stuff would drag in lots of glibc's locale handling
stuff which is completely incompatible with Microsoft's C library's
locale handling anyway. Also, I am not sure whether the GNU regex code
is prepared to handle a two-byte wchar_t, or does it assume that
wchar_t is int as it is on Linux? Hmm, actually there is lots of
sizeof(wchar_t) in glibc, so maybe it *is* prepared? Maybe
later... But anyway, it would presumably mean we should have not just
the regex functionality but a larger subset of glibc that would
include all locale, ctype, wchar, mbs, etc stuff, presumably ending up
with a very large part of glibc (not the system calls,
obviously). Indeed, something to save for later, or never...

--Tor Lillqvist <[email protected]>, <[email protected]>