Hi,

> BUT - for compiler '\355' equals to -19, so the macro looks truly
> like this:
> 
> ((_ctype_ + 1)[-19] & (_P|_U|_L|_N|_B))
>              ^^^^^^^^
> I wouldn't expect this is what we want.
> 
> Every subject to any <ctype.h> macro have to be casted to unsigned
> char to make it work correctly.

I think we should follow Peter Seebachs advice and make 
char unsigned by default to eliminate the casts. This would
help a lot with other software as well, like X and its bitmaps...

> Should I send-pr it or am I just missing something ?

I'd say go for it!

[make locale work]
> I have noticed one unclosed send-pr relative to this problem
> (lib/2702, 1.2_BETA system). Oh, and lib/2309 too, it complains
> about missing manpage to setlocale(3), it's from April 6, 1996.

I, and some others, have complained about this a couple of times,
but noone ever commented on this. I'd say, if you have the time,
go for it and do a send-pr. That way others can integrate it,
and I guess core wants it as well, but doesn't have the time to
deal with it. I guess a proposal on how you want to implement
it would be the best solution. If noone objects to this, then
it's really likely to have the solution integrated. That way,
you don't waste any time reorganizing/rewriting code later on.

  Guenther