Re: [rfc/rft] Update remaining targets to value-based unwinding

["Ulrich Weigand" <uweigand@de.ibm.com>]

DJ Delorie wrote:
> > > m32c (two multilibs):
> > > 
> > > # of expected passes            17068
> > > # of unexpected failures        2758
> > 
> > In particular, this result looks a bit worrisome ...
> 
> It's bad.  The m32c is a weird chip (two stack pointers, 24 bit
> addresses), it will probably need Jim or Kevin to massage your patch
> to get it to work.

Ah, I see I made an obvious mistake in the m32c conversion.

Would you mind testing again with the following additional patch?

Thanks,
Ulrich

diff -urNp gdb-orig/gdb/m32c-tdep.c gdb-head/gdb/m32c-tdep.c
--- gdb-orig/gdb/m32c-tdep.c	2008-05-20 00:41:46.000000000 +0200
+++ gdb-head/gdb/m32c-tdep.c	2008-05-20 00:42:23.000000000 +0200
@@ -1919,7 +1919,7 @@ m32c_prev_register (struct frame_info *t
      return a description of the stack slot holding it.  */
   if (p->reg_offset[regnum] != 1)
     return frame_unwind_got_memory (this_frame, regnum,
-                                    p->reg_offset[regnum]);
+                                    frame_base + p->reg_offset[regnum]);
 
   /* Otherwise, presume we haven't changed the value of this
      register, and get it from the next frame.  */


-- 
  Dr. Ulrich Weigand
  GNU Toolchain for Linux on System z and Cell BE
  Ulrich.Weigand@de.ibm.com