Kevin Lawton wrote:
> I wanted to have one set of types. I hate having a mixture
> of Bit32u and u32. It's confusing.
Well, yes, I wanted to have one set of types as well: only the Bit...
ones ;-) Those are already in use throughout the user space app, and
at most places in the kernel as well. The only reason for using u32 etc.
might be inside host-linux.c when calling a Linux routine that requires
u32 as argument type. (But inside host-linux.c we have to use Linux-specific
types anyway, so that's no big deal ...)
Everywhere else there's really no point in requiring Linux headers IMO.
> Since Linux defines
> this type, and conditionally types with leading double underscore
> (depending if you're compile user or kernel code), I had to
> add a little #ifdef code to make it work.
Well, the point is that it doesn't work at all if you aren't compiling
under Linux in the first place. It just strikes me as odd to require
Linux headers *everywhere* for something as trivial as this, especially
considering that we define our own types *anyway*. If you just like the
name u32 better than Bit32u, I'd still prefer a 'typedef Bit32u u32' or
something to the current hack ...
> > Another CVS-specific problem: it is impossible to rename directories
> > in CVS, so that in order to rename the main directory to plex86 we'll
> > probably have to create a new repository anyway ...
>
> Yeah, just as well.
You want to do it yourself, or should I do it?
Bye,
Ulrich
--
Ulrich Weigand,
IMMD 1, Universitaet Erlangen-Nuernberg,
Martensstr. 3, D-91058 Erlangen, Phone: +49 9131 85-27688
