This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [RFC] h8300 "info registers" fix
> Date: Wed, 13 Sep 2006 13:49:26 -0700
> From: "Michael Snyder" <Michael.Snyder@palmsource.com>
>
> Daniel:
> >On Wed, Sep 13, 2006 at 01:20:09PM -0700, Michael Snyder wrote:
> >> Grumble -- is the regcache kept in host order, or target order?
> >
> >Target order.
> >
> >[Which is apparently a bit weird; most debug interfaces I've seen
> >lately use host order.]
>
> That's right -- it's a relic of the fact that the original
> "register cache" was the register packet itself, which is
> generated on the target side. The target doesn't know anything
> about the host, so there's no way that can be in host order.
>
> Anyway, in that case, the gdb_byte[4] approach is more likely
> to be correct, eh?
Yes, gdb_byte is the type we use for target byte buffers.
> I think I understand the problem now -- the pseudo-register is
> only one byte, so it's "natural" to call pseudo_register_read
> with a one byte buffer. But the physical register is four bytes,
> so you have to have a four byte buffer to read it.
>
> Wonder why it ever worked? ;-/
The fact that almost all variables on a 32-bit machines are
word-aligned?
Mark