This is the mail archive of the
gdb-patches@sources.redhat.com
mailing list for the GDB project.
Re: [RFC] Trivialize objc-lang.c FETCH_ARGUMENT
- From: Daniel Jacobowitz <drow at mvista dot com>
- To: Michael Snyder <msnyder at redhat dot com>
- Cc: Adam Fedor <fedor at doc dot com>,GDB Patches <gdb-patches at sources dot redhat dot com>
- Date: Tue, 22 Apr 2003 22:14:00 -0400
- Subject: Re: [RFC] Trivialize objc-lang.c FETCH_ARGUMENT
- References: <5D31A204-752F-11D7-A2CD-000A277AC1A4@doc.com> <3EA5F51A.8CC11F12@redhat.com>
On Tue, Apr 22, 2003 at 07:06:18PM -0700, Michael Snyder wrote:
> Adam Fedor wrote:
> >
> > On Tuesday, April 22, 2003, at 07:41 PM, Michael Snyder wrote:
> >
> > > Adam Fedor wrote:
> > >>
> > >> Well, I thought I'd at least try to see if this patch would be ok.
> > >> Here
> > >> I replace the arcitecture dependant FETCH_ARGUMENT with a trivial
> > >> implementation that does nothing. This would allow objc-lang.o to be
> > >> linked into gdb and I commit most or all of the remaining Objective-C
> > >> patches. Then I could work on fixing FETCH_ARGUMENT at my leasure...
> > >
> > > That seems reasonable. By the way, if there's been an ongoing
> > > discussion,
> > > I haven't followed it. Why is it that you need to do this in an
> > > architecture-dependent way? GDB should have enough debug info to
> > > do this cleanly, shouldn't it?
> > >
> > It's possible, although these particular functions are in the Apple
> > runtime and highly optimized, possibly in assembly. I'm not sure if the
> > information is available. I'll have to look at it more, but I couldn't
> > even test the changes since it only works on MacOSX/Darwin and GNU gdb
> > doesn't compile on Darwin.
>
> Yeah, if I remember from my days at NeXT, the purpose of this was
> in debugging something that was not really debuggable.
>
> So some sort of compromise would seem to be in order.
>
> If this function does not need to be general (it doesn't seem to
> be used anywhere else), perhaps the apple folk can agree to some
> sort of a specified API for obtaining the arguments they need.
> Or perhaps they no longer need such radical optimization, and
> the functions can be written in C.
The functions are written in C; they just don't necessarily have debug
info. If you search for FETCH_POINTER_ARGUMENT in the archives (about
a month ago?) you'll see that I need something similar for C++; so I
support just getting this method properly defined and into the
architecure where it's feasible to.
--
Daniel Jacobowitz
MontaVista Software Debian GNU/Linux Developer