> On Sun, Dec 09, 2001 at 11:57:35AM -0800, Jason R Thorpe wrote:
> 
>  > I am planning to fix the PowerPC ptrace(2) (while still maintaining
>  > binary compatibility, possible with some other ptrace(2) infrastructure
>  > changes I made last week) to provide register info like so:
>  > > struct reg {
>  > register_t	r_fixreg[32];
>  > register_t	r_srr0;		/* PC */
>  > register_t	r_srr1;		/* MSR */
>  > register_t	r_lr;
>  > register_t	r_ctr;
>  > register_t	r_mq;		/* unused on PowerPC */
>  > int		r_cr;
>  > int		r_xer;
>  > };


The proposal to make the core file layout map the ptrace buffer, I think 
is a good move.  Many targets use the same functions to unpack both.

I suspect that expanding the size of the ptrace() buffer, however, is a 
potential ``you loose'' situtation.  Consider what happens if someone 
tries to debug a process using an existing (eg 1.5.2) GDB binary.  As 
far as I know, unless you play around with syscall numbers, that binary 
will get the tail end of its ptrace buffer smashed by the request.

The extra registers would be useful though!

enjoy,
Andrew