ptrace for mips n32 and n64

[Daniel Jacobowitz]

On Fri, Mar 14, 2003 at 10:30:14AM -0700, Kevin Buettner wrote:
> On Mar 14, 11:57am, Daniel Jacobowitz wrote:
> 
> > I still don't really like it - it will make porting native GDB a real
> > pain, 
> 
> Why will it be a pain?  GDB has the macros PTRACE_XFER_TYPE and
> PTRACE_ARG3_TYPE to make this sort of thing easier.
> 
> FWIW, I don't recall any n32 related problems related to ptrace()
> when I did the rda port.  (I must admit that I found this kind of
> surprising.)

I take it back; I misspoke.  We used to have a lot of PTRACE_XFER_TYPE
and PTRACE_ARG3_TYPE missing, and I remember cleaning them up (for
mips-linux?  maybe, but I don't remember why now.)

It will be a pain for gdbserver, but that's my problem.  I don't have
target-specific macros right now and I've managed to avoid them thus
far.  Adding one won't kill me, and that segment could use a cleanup.

> > which is probably why Kevin only did rda.
> 
> The reason I did a port for rda and not native gdb is because that is
> what I was assigned to do, not because it was going to be a pain.

Sorry.

> The thing which I think is going to be a problem is the fact that
> there are three libthread_db libraries, one for each of the ABIs
> to contend with.  If that weren't the case, I think a single n64
> gdb could suffice for debugging applications using any of the three
> ABIs.
> 
> I couldn't think of a nice way to solve this, so, for rda, we have
> three different rda binaries, one for each ABI.

Yes.  This is one of the fundamental problems with libthread_db.  I
expect we'll need to do the same thing for native GDB and for gdbserver
also.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer