collation tables

[Dmitry Timoshkov]

"Vitaly Lipatov" <lav at altlinux.ru> wrote:

> > Could you post a couple of examples what exactly differences
> > do you observe? And what hacks do you propose?
> There is attached the sorted (ordered) symbols from codepage 
> cp1251. I get it from output 3.c program runned in win2003, wine 
> and wine with collate from crossover.

Yes, there are some difference, but I wouldn't call them fatal.
All national characters are sorted correctly. If 1C does a binary
sort using lstrcmpA and does comparisons with hard coded results
then it's a really broken app. In order to "fix" the app we have
to patch the whole unicode range, not just cyrillic char map.

> I guess do hack for 8bit stpcmp functions (non-Unicode) for 
> cp1251 only.

I'm afraid that hack could not be accepted under any circumstances
for inclusion into the Wine tree.

My guess is that we have to maintain collation weights for the whole
unicode range (or at least some unicode ranges) on our own, and base
a patch for unicode.org collation tables on that. That's a huge and
very tedious work, I know that, since I'm the author of the collation
table patch in Crossover.

-- 
Dmitry.