Kris Warkentin <kewarken@qnx.com> wrote:

> > Any reason why you can't use the generic register support in
> > i386-tdep.c?
>
> Actually, I take advantage of the i387 generics for the floating point
> regs since our kernel just uses the opcode to push them in the normal
> order.  Unfortunately though, our gpregs are in a different order and
> not all of them are there so we have some unique mappings.

QNX NTO is in no way special about this.  I was thinking about something like the attached patch.  Could you test that one?  This approach simply reduces the amount of code needed.

The patch also contains a unrelated change to the shared library code.  It also removes a unneeded function.  But it'd be nice if you could test that I didn't mess things up.

Mark