[RFA/RFC] Add i386_cleanup_dregs at program exit.

[RFA/RFC] Add i386_cleanup_dregs at program exit.

[Pierre Muller]

   A rather long thread started by my
description of a failure to catch hardware watchpoints
on the second run of a linux i386 executable when debugging with 5.1 version.

See
http://sources.redhat.com/ml/gdb-patches/2001-11/msg00613.html
that contains a description of how to reproduce this bug.

The following patch does fix the problem and 
has the further advantage of being multi-target,
so it might well fix the same bug on FreeBSD for example 
(I didn't test if this bug is also present in FreeBSD).
But it isn't multi-arch yet.

The patch simply calls i386_cleanup_dregs
in breakpoint_init_inferior if context is inf_exited.

I have no idea if other processor might need a similar 
cleanup of debug registers, but if there are,
simply setting the HARDWARE_REGISTER_CLEANUP
macro should be enough.

Andrew, how can this be multi-arched ??
Should I add something like 
#ifndef HARDWARE_REGISTER_CLEANUP_P
#define HARDWARE_REGISTER_CLEANUP_P 0
#endif 
#ifndef HARDWARE_REGISTER_CLEANUP
#define HARDWARE_REGISTER_CLEANUP {}
#endif 
and replace the patch in breakpoint.c by
  if (HARDWARE_REGISTER_CLEANUP_P
      && (context == inf_exited))
       HARDWARE_REGISTER_CLEANUP; 

This removes the #ifdef
but full multi-arch probably will require 
to add two new fields to the gdbarch vector ?


  
2002-01-09  Pierre Muller <muller@ics.u-strasbg.fr>

	* breakpoint.c (breakpoint_init_inferior): Call
	HARDWARE_REGISTER_CLEANUP macro if defined
	if context is inf_exited.
	* config/i386/nm-i386.h: Define HARDWARE_REGISTER_CLEANUP
	macro to call i386_cleanup_dregs().

Index: breakpoint.c
===================================================================
RCS file: /cvs/src/src/gdb/breakpoint.c,v
retrieving revision 1.57
diff -u -r1.57 breakpoint.c
--- breakpoint.c        2001/11/11 16:39:59     1.57
+++ breakpoint.c        2002/01/08 23:44:00
@@ -1487,6 +1487,10 @@
        warning ("You must reinsert them explicitly.");
        warning_needed = 0;
      }
+#ifdef HARDWARE_REGISTER_CLEANUP
+  if (context == inf_exited)
+     HARDWARE_REGISTER_CLEANUP;
+#endif
  }

  /* breakpoint_here_p (PC) returns non-zero if an enabled breakpoint
Index: config/i386/nm-i386.h
===================================================================
RCS file: /cvs/src/src/gdb/config/i386/nm-i386.h,v
retrieving revision 1.2
diff -u -r1.2 nm-i386.h
--- nm-i386.h   2001/03/23 16:17:45     1.2
+++ nm-i386.h   2002/01/08 23:44:01
@@ -115,6 +115,8 @@

  #define DECR_PC_AFTER_HW_BREAK 0

+#define HARDWARE_REGISTER_CLEANUP i386_cleanup_dregs ()
+
  #endif /* I386_USE_GENERIC_WATCHPOINTS */

  #endif /* NM_I386_H */




Pierre Muller
Institut Charles Sadron
6,rue Boussingault
F 67083 STRASBOURG CEDEX (France)
mailto:muller@ics.u-strasbg.fr
Phone : (33)-3-88-41-40-07  Fax : (33)-3-88-41-40-99

Re: [RFA/RFC] Add i386_cleanup_dregs at program exit.

[Eli Zaretskii]

On Wed, 9 Jan 2002, Pierre Muller wrote:

> The patch simply calls i386_cleanup_dregs
> in breakpoint_init_inferior if context is inf_exited.

Is this really necessary?  Why do we need to propagate processor-specific 
issues into breakpoint.c?

I thought there was a simpler solution back when we discussed this.  
Something that would leave this issue where it belongs: in the 
x86-specific files.

> I have no idea if other processor might need a similar 
> cleanup of debug registers

They probably don't, since we didn't hear anything from their users.

Re: [RFA/RFC] Add i386_cleanup_dregs at program exit.

[Pierre Muller]

At 10:20 09/01/2002 , Eli Zaretskii a écrit:

>On Wed, 9 Jan 2002, Pierre Muller wrote:
>
> > The patch simply calls i386_cleanup_dregs
> > in breakpoint_init_inferior if context is inf_exited.
>
>Is this really necessary?  Why do we need to propagate processor-specific 
>issues into breakpoint.c?

Because there is no i386 common code that is called at program exit,
so otherwise this would need a much bigger change, with adding processor specific
mourn_inferior functions for OS that have a generic one
like linux.


>I thought there was a simpler solution back when we discussed this.  
>Something that would leave this issue where it belongs: in the 
>x86-specific files.
>
> > I have no idea if other processor might need a similar 
> > cleanup of debug registers
>
>They probably don't, since we didn't hear anything from their users.

But they might wake up later....


Pierre Muller
Institut Charles Sadron
6,rue Boussingault
F 67083 STRASBOURG CEDEX (France)
mailto:muller@ics.u-strasbg.fr
Phone : (33)-3-88-41-40-07  Fax : (33)-3-88-41-40-99

Re: [RFA/RFC] Add i386_cleanup_dregs at program exit.

[Eli Zaretskii]

On Wed, 9 Jan 2002, Pierre Muller wrote:

> >Is this really necessary?  Why do we need to propagate processor-specific 
> >issues into breakpoint.c?
> 
> Because there is no i386 common code that is called at program exit,

If we really need that, we could add one.  There's nothing wrong IMHO 
with having a target-specific mourne-inferior function.