Test version of glibc 2.3.5-10

[Petr Salinger]

> > Yes, I tried to achieve no ABI changes at all, 
> > leave glibc structures identical to previous. 
> > Advantage is no need to rebuild anything,
> > struct stat doesn't contain any 64 bit fields
> > (no unexpected warnings for printf format string).
> > 
> > Advantage of your approach is that it doesn't truncate 
> > fields like f_syncwrites on 4GB, current spare fields
> > will be available to user space. 
> 
> Well I don't really know what to do. Those functions are not POSIX 
> functions, so there is no rules. The old version of the statfs structure 
> is near to the GNU/Linux one, the new version is the default FreeBSD 5.x 
> or 6.x one.

Well, new structure might break something, it might be passed between 
libs. At least for initial upload to unreleased it will be better to not 
change these structures.

It might be changed later:

- just before upload to debian.org
- with glibc bump from 2.3.x to 2.4.x
- ...

Alternatively, leave struct stat consistent with 2.3.0,
change only struct stat64 to current kernel structure.
It would vote for this, because 
it will be consistent with spirit of xxx/xxx64.

Petr

P.S.
Rebuild with my patch finished, it looks ok.