>  So the first step is to define an interface for getting at "extra"
>    registers such as Altivec and SSE - a PT_GETOTHERREGS would probably
>    work, but will lose as soon as we get Altivec2 registers or something
>    - and then do a little back-end glue in the PowerPC process_machdep.c
>    to recognize it.

SNAFU. PT_GETOTHEROTHERREGS of course.

> couldn't we use the "int data" param of ptrace(2) to allow the new
> PT_[GS]ETOTHERREGS to have an additional argument?  that would allow
> fairly extensive expansion :-)

``int data', what ``int data'' parameter?  Oh that one.  Hmmm, interesting.

Same problem occures with core.  The convention is .reg (gen) & .reg2 
(fp) but after that all bets appear to be off.  Is there a standard

	Andrew

(There is apparently also a (i386? XXX kernel? specific) .reg-xfp 
section kicking around but I have no idea where that came from or if 
anything even uses it.)