Re: [RFA] Altivec ABI patches

[Elena Zannoni <ezannoni at redhat dot com>]

Jim Blandy writes:
 > 
 > I don't see anything here that prevents vectors from being coerced to
 > pointers to their first element.  Isn't there some change needed in
 > evaluate_subexp_with_coercion?


You mean value_arg_coerce()?  There is no arithmetic you can do with
these types.  Anyway, you are right I forgot that. My tests didn't
catch it because I was doing the call by hand after having done a step
into and finish from the function. The registers still contained the
correct values from the gcc-initiated call. Moving the gdb-initiated
call to before the gcc-initiated one exposed the problem.

I'll make sure the registers are trashed between calls.

Elena