Re: gdb/854: gdb HC11/HC12 broken due to regcache

[Andrew Cagney <ac131313 at redhat dot com>]

The following reply was made to PR gdb/854; it has been noted by GNATS.

From: Andrew Cagney <ac131313@redhat.com>
To: Stephane.Carrez@nerim.fr
Cc: gdb-gnats@sources.redhat.com
Subject: Re: gdb/854: gdb HC11/HC12 broken due to regcache
Date: Thu, 28 Nov 2002 17:40:27 -0500

 > gdb for m6811-elf is broken due to the new regcache.
 > 
 > The regcache only saves the real hard register and does not
 > save the HC11/HC12 memory registers.  The frame pointer is
 > also a memory register and it is not saved by the regcache.
 > This results in wrong behavior when using backtrace and
 > when gdb call a target function (because the generic_dummy
 > frame only saves the regcache, and the HC11/HC12 memory
 > registers {and in particular the frame} are not saved).
 > 
 >>How-To-Repeat:
 > 
 > Run gdb validation and in particular
 > testsuite/gdb.base/callfuncs.exp
 > 
 >>Fix:
 > 
 > Must save the HC11/HC12 memory registers
 
 I knew it!  The regcache code should be able to accomodate this vis:
 
    /* Copy over any registers (identified by their membership in the
       save_reggroup) and mark them as valid.  The full [0
       .. NUM_REGS+NUM_PSEUDO_REGS) range is checked since some
       architectures need to save/restore `cooked' registers that live
       in memory.  */
 
 You should be able to define a local m68hc11_register_reggroup_p() that 
 is something like:
 
 	if (reggroup == save_reggroup
 	    || reggroup == restore_reggroup)
 	  return one when it is one of th registers that need saving/restoring;
 	return default_register_reggroup_p (reggroup, regnum);
 
 while you're at it you may want to refine some of the other reggroups so 
   that they better reflect you're architectures reality (see the i386 
 for an example).