On Wed, 9 Oct 2002, Brent Dax wrote:
> Andy Dougherty:
> # Overall, it continues to look very good. Unfortunately, the
> # compact-looking
> #
> # ch=va_arg((va_list)obj->data, char);
> #
> # gave Sun's compiler indigestion. I had to split it up into
> # the much more pedestrian
> #
> # va_list arg;
> # arg = (va_list) obj->data;
> # ch = va_arg(arg, char);
>
> Can you try putting the first code back in, but change the object call
> to be like this:
>
> ((va_list)(obj->data))
>
> ? I'm not sure yet, but this *may* be causing your other problems.
Duh, of course, va_list() can have side effects. My fault. Sorry.
Alas, your suggestion doesn't compile either, but I'll poke and prod
and see what I can do to make it compile.
Hmm. Just removing the (va_list) cast makes both Sun's compiler and
gcc happy. (That's because va_list is ultimately typedef'd to void *
anyway under Solaris, and probably most other Unix systems. I don't
know about elsewhere, so this could be tricky.)
Offhand, I'd say simply remove the (va_list) casts. If that proves a
problem for some platform, I'll poke and prod around a bit and see if I
can fix things up. I may end up doing something vaguely like
va_list arg;
arg = (va_list) obj->data;
ch = va_arg(arg, char);
obj->data = (void *) arg;
everywhere, though just removing the casts is easier if it works :-).
--
Andy Dougherty [EMAIL PROTECTED]
