Re: [patch] Translation of mn10300 debugging symbols to gdb's internalregister numbering scheme

Re: [patch] Translation of mn10300 debugging symbols to gdb's internalregister numbering scheme

[Michael Snyder]

Matt Hiller wrote:
> 
>         At http://gcc.gnu.org/ml/gcc-patches/2001-02/msg01491.html , Alex
> Oliva describes an inconsistency in the register numbering for mn10300
> between gcc and the other tools. Matushita has developed debugging tools
> based on the numbering gcc uses, so we are obliged to alter the behavior
> of gdb instead. The following patch has gdb translate debugging register
> numbers to gdb REGNUMs.
> 
>         I do not have write access to this repository; my patch, if
> acceptable, will have to be committed by someone else.

Matt, 

I think you have found the right approach, but not quite the right
implementation.  The DWARF_REG_TO_REGNUM macro has been multi-arched.
That means that instead of defining it as a macro, you should implement
it as a function in mn10300-tdep.c, and then set the appropriate 
gdbarch function vector to point to your function.

While you are at it, I would recommend you go ahead and use the
same function for dwarf2_reg_to_regnum.  Might as well cover
all bases.

				Thanks,
				Michael


> 
> --
> 
> Matt Hiller
> GCC Engineer, Red Hat, Inc., Sunnyvale office
> hiller@redhat.com
> 
> 2001-02-27  Matt Hiller  <hiller@redhat.com>
> 
>         * config/mn10300/tm-mn10300.h (DWARF_REG_TO_REGNUM): New macro.
>         (STAB_REG_TO_REGNUM): New macro.
> 
> Index: config/mn10300/tm-mn10300.h
> ===================================================================
> RCS file: /cvs/src/src/gdb/config/mn10300/tm-mn10300.h,v
> retrieving revision 1.4
> diff -u -p -r1.4 tm-mn10300.h
> --- config/mn10300/tm-mn10300.h 2000/08/12 03:28:42     1.4
> +++ config/mn10300/tm-mn10300.h 2001/02/27 21:33:46
> @@ -167,3 +167,15 @@ extern use_struct_convention_fn mn10300_
>  extern void mn10300_virtual_frame_pointer (CORE_ADDR, long *, long *);
>  #define TARGET_VIRTUAL_FRAME_POINTER(PC, REGP, OFFP) \
>         mn10300_virtual_frame_pointer ((PC), (REGP), (OFFP))
> +
> +/* Convert a DWARF register number to a gdb REGNUM.  */
> +#define DWARF_REG_TO_REGNUM(num) \
> +  (((num) <= 7) ? (num) \
> +   : (num == 8) ? (-1)   /* ap, which won't appear in debugging info.  */ \
> +   : (num == 9) ? (SP_REGNUM) \
> +   : (10 <= num && num <= 17) ? ((num - 10) + E0_REGNUM) \
> +   : (-1))   /* program counter, etc., none of which appear in debugging
> +               info.  */
> +
> +/* Convert a STAB register number to a gdb REGNUM.  */
> +#define STAB_REG_TO_REGNUM(num) DWARF_REG_TO_REGNUM (num)

Re: [patch] Translation of mn10300 debugging symbols to gdb's internalregister numbering scheme

[Matt Hiller]

On Tue, 27 Feb 2001, Michael Snyder wrote:
> 
> Matt, 
> 
> I think you have found the right approach, but not quite the right
> implementation.  The DWARF_REG_TO_REGNUM macro has been multi-arched.
> That means that instead of defining it as a macro, you should implement
> it as a function in mn10300-tdep.c, and then set the appropriate 
> gdbarch function vector to point to your function.
> 
> While you are at it, I would recommend you go ahead and use the
> same function for dwarf2_reg_to_regnum.  Might as well cover
> all bases.

	Okay. Another draft follows:

> Matt Hiller wrote:
> > 
> >         At http://gcc.gnu.org/ml/gcc-patches/2001-02/msg01491.html , Alex
> > Oliva describes an inconsistency in the register numbering for mn10300
> > between gcc and the other tools. Matushita has developed debugging tools
> > based on the numbering gcc uses, so we are obliged to alter the behavior
> > of gdb instead. The following patch has gdb translate debugging register
> > numbers to gdb REGNUMs.
> > 
> >         I do not have write access to this repository; my patch, if
> > acceptable, will have to be committed by someone else.

2001-02-27  Matt Hiller  <hiller@redhat.com>

	* mn10300-tdep.c (mn10300_stab_reg_to_regnum): New function.
	(mn10300_gdbarch_init): Set appropriate elements of gdbarch to
	mn10300_stab_reg_to_regnum.
 
Index: mn10300-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/mn10300-tdep.c,v
retrieving revision 1.6
diff -u -p -r1.6 mn10300-tdep.c
--- mn10300-tdep.c	2001/02/08 06:03:53	1.6
+++ mn10300-tdep.c	2001/02/28 00:01:35
@@ -939,6 +939,25 @@ mn10300_dump_tdep (struct gdbarch *curre
 		      tdep->am33_mode);
 }
 
+/* Convert a dbx stab register number to a gdb REGNUM.  Also fine for
+   dwarf and dwarf2.  */
+
+static int
+mn10300_stab_reg_to_regnum (int num)
+{
+  if (num <= 7)
+    return num;
+  else if (num == 8)
+    return -1;  /* ap, which won't appear in debugging info.  */
+  else if (num == 9)
+    return SP_REGNUM;
+  else if (10 <= num && num <= 17)
+    return (num - 10) + E0_REGNUM;
+  else
+    return -1;  /* program counter, etc., none of which appear in debugging
+		   info.  */
+}
+
 static struct gdbarch *
 mn10300_gdbarch_init (struct gdbarch_info info,
 		      struct gdbarch_list *arches)
@@ -999,6 +1018,9 @@ mn10300_gdbarch_init (struct gdbarch_inf
   set_gdbarch_save_dummy_frame_tos (gdbarch, generic_save_dummy_frame_tos);
   set_gdbarch_num_regs (gdbarch, num_regs);
   set_gdbarch_do_registers_info (gdbarch, mn10300_do_registers_info);
+  set_gdbarch_stab_reg_to_regnum (gdbarch, mn10300_stab_reg_to_regnum);
+  set_gdbarch_dwarf_reg_to_regnum (gdbarch, mn10300_stab_reg_to_regnum);
+  set_gdbarch_dwarf2_reg_to_regnum (gdbarch, mn10300_stab_reg_to_regnum);
 
   tdep->am33_mode = am33_mode;