expand_expr sign extension problem.

[Robert Lipe]

> > 	#ifndef __GNUC__
> > 	#define INT_ONLY_FIELDS
> > 	#endif
> 
> But then you loose the ability for a port to say its non-GCC compiler does
> support enum bitfields.  I'll defer that though to those that use such a
> compiler.

But that isn't precisely what INT_ONLY_FIELDS explictly circumvents.  If
you have a compiler that does support enum bitfields in a way that the
standards allow them to (signed instead of unsigned) that is also what
INT_ONLY_FIELDS gets around.   It implictly circumvents this behaviour
by sandbagging extra bits before the sign bits...

> > I've just thought of a good reason why stages 2 & 3 might 
> > need to be built with INT_ONLY_FIELDS defined. What if the
> > host debugger can't handle enumerated bit fields!
> 
> Or you use gdb as the host debugger, and don't use COFF debug format.

I can't quite parse that, Michael.  Are you saying that COFF debug
format doesn't support this or that GDB w/ COFF debugging doesn't
support this or that GDB with, say, stabs, doesn't support this?

RJL