m68k_register_virtual_type

m68k_register_virtual_type

[Andreas Schwab]

[ Seems like this mail didn't come through although I'm sure it left my
notebook. ]

The m68k sr, fpcontrol and fpstatus registers don't contain addresses, so
they should be treated like any other integer register.  Committed.

Andreas.

2002-11-22  Andreas Schwab  <schwab@suse.de>

	* m68k-tdep.c (m68k_register_virtual_type): Return int for SR, FPC
	and FPS registers.

Index: gdb/m68k-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/m68k-tdep.c,v
retrieving revision 1.27
retrieving revision 1.28
diff -u -p -a -u -p -a -r1.27 -r1.28
--- gdb/m68k-tdep.c	18 Nov 2002 22:19:28 -0000	1.27
+++ gdb/m68k-tdep.c	22 Nov 2002 00:02:03 -0000	1.28
@@ -148,20 +148,25 @@ m68k_register_virtual_size (int regnum)
   return (((unsigned) (regnum) - FP0_REGNUM) < 8 ? 12 : 4);
 }
 
-/* Return the GDB type object for the "standard" data type of data 
-   in register N.  This should be int for D0-D7, long double for FP0-FP7,
-   and void pointer for all others (A0-A7, PC, SR, FPCONTROL etc).
-   Note, for registers which contain addresses return pointer to void, 
-   not pointer to char, because we don't want to attempt to print 
-   the string after printing the address.  */
+/* Return the GDB type object for the "standard" data type of data in
+   register N.  This should be int for D0-D7, SR, FPCONTROL and
+   FPSTATUS, long double for FP0-FP7, and void pointer for all others
+   (A0-A7, PC, FPIADDR).  Note, for registers which contain
+   addresses return pointer to void, not pointer to char, because we
+   don't want to attempt to print the string after printing the
+   address.  */
 
 static struct type *
 m68k_register_virtual_type (int regnum)
 {
-  if ((unsigned) regnum >= E_FPC_REGNUM)
+  if (regnum == E_FPI_REGNUM)
     return lookup_pointer_type (builtin_type_void);
+  else if ((unsigned) regnum >= E_FPC_REGNUM)
+    return builtin_type_int;
   else if ((unsigned) regnum >= FP0_REGNUM)
     return builtin_type_long_double;
+  else if (regnum == PS_REGNUM)
+    return builtin_type_int;
   else if ((unsigned) regnum >= A0_REGNUM)
     return lookup_pointer_type (builtin_type_void);
   else

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: m68k_register_virtual_type

[Andrew Cagney]

Just FYI,

>  static struct type *
>  m68k_register_virtual_type (int regnum)
>  {
> -  if ((unsigned) regnum >= E_FPC_REGNUM)
> +  if (regnum == E_FPI_REGNUM)
>      return lookup_pointer_type (builtin_type_void);

There are now:
	builtin_type_void_data_ptr
	builtin_type_void_code_ptr
available.


> +  else if ((unsigned) regnum >= E_FPC_REGNUM)
> +    return builtin_type_int;

builtin_type_int32 is better

>    else if ((unsigned) regnum >= FP0_REGNUM)
>      return builtin_type_long_double;
> +  else if (regnum == PS_REGNUM)
> +    return builtin_type_int

Ditto.

>    else if ((unsigned) regnum >= A0_REGNUM)
>      return lookup_pointer_type (builtin_type_void);
>    else

Andrew

Re: m68k_register_virtual_type

[Andreas Schwab]

Andrew Cagney <ac131313@redhat.com> writes:

|> Just FYI,
|> 
|> >  static struct type *
|> >  m68k_register_virtual_type (int regnum)
|> >  {
|> > -  if ((unsigned) regnum >= E_FPC_REGNUM)
|> > +  if (regnum == E_FPI_REGNUM)
|> >      return lookup_pointer_type (builtin_type_void);
|> 
|> There are now:
|> 	builtin_type_void_data_ptr
|> 	builtin_type_void_code_ptr
|> available.

Thanks, I'll change that (it's actually builtin_type_void_func_ptr).

|> > +  else if ((unsigned) regnum >= E_FPC_REGNUM)
|> > +    return builtin_type_int;
|> 
|> builtin_type_int32 is better
|> 
|> >    else if ((unsigned) regnum >= FP0_REGNUM)
|> >      return builtin_type_long_double;
|> > +  else if (regnum == PS_REGNUM)
|> > +    return builtin_type_int
|> 
|> Ditto.

What about the default case for the d0-d7 registers?  Should it be changed
to builtin_type_int32 as well?

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: m68k_register_virtual_type

[Andrew Cagney]

> |> > +  else if ((unsigned) regnum >= E_FPC_REGNUM)
> |> > +    return builtin_type_int;
> |> 
> |> builtin_type_int32 is better
> |> 
> |> >    else if ((unsigned) regnum >= FP0_REGNUM)
> |> >      return builtin_type_long_double;
> |> > +  else if (regnum == PS_REGNUM)
> |> > +    return builtin_type_int
> |> 
> |> Ditto.
> 
> What about the default case for the d0-d7 registers?  Should it be changed
> to builtin_type_int32 as well?

Er, yes.  (And eventually long_double to builtin_type_m68881_ext but the 
last one might have fallout related to register conversions).

Andrew

Re: m68k_register_virtual_type

[Andreas Schwab]

Andrew Cagney <ac131313@redhat.com> writes:

|> > |> > +  else if ((unsigned) regnum >= E_FPC_REGNUM)
|> > |> > +    return builtin_type_int;
|> > |> |> builtin_type_int32 is better
|> > |> |> >    else if ((unsigned) regnum >= FP0_REGNUM)
|> > |> >      return builtin_type_long_double;
|> > |> > +  else if (regnum == PS_REGNUM)
|> > |> > +    return builtin_type_int
|> > |> |> Ditto.
|> > What about the default case for the d0-d7 registers?  Should it be
|> > changed
|> > to builtin_type_int32 as well?
|> 
|> Er, yes.  (And eventually long_double to builtin_type_m68881_ext but the
|> last one might have fallout related to register conversions).

What kind of fallout could that be?  Some quick tests didn't reveal
anything obvious.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: m68k_register_virtual_type

[Andreas Schwab]

Andrew Cagney <ac131313@redhat.com> writes:

|> Er, yes.  (And eventually long_double to builtin_type_m68881_ext but the
|> last one might have fallout related to register conversions).

I have now committed this patch.

Andreas.

2002-11-23  Andreas Schwab  <schwab@suse.de>

	* m68k-tdep.c (m68k_register_virtual_type): Use architecture
	invariant return values.

--- m68k-tdep.c.~1.28.~	2002-11-23 00:53:11.000000000 +0100
+++ m68k-tdep.c	2002-11-23 00:53:32.000000000 +0100
@@ -159,18 +159,19 @@ m68k_register_virtual_size (int regnum)
 static struct type *
 m68k_register_virtual_type (int regnum)
 {
-  if (regnum == E_FPI_REGNUM)
-    return lookup_pointer_type (builtin_type_void);
-  else if ((unsigned) regnum >= E_FPC_REGNUM)
-    return builtin_type_int;
-  else if ((unsigned) regnum >= FP0_REGNUM)
-    return builtin_type_long_double;
-  else if (regnum == PS_REGNUM)
-    return builtin_type_int;
-  else if ((unsigned) regnum >= A0_REGNUM)
-    return lookup_pointer_type (builtin_type_void);
-  else
-    return builtin_type_int;
+  if (regnum >= FP0_REGNUM && regnum <= FP0_REGNUM + 7)
+    return builtin_type_m68881_ext;
+
+  if (regnum == E_FPI_REGNUM || regnum == PC_REGNUM)
+    return builtin_type_void_func_ptr;
+
+  if (regnum == E_FPC_REGNUM || regnum == E_FPS_REGNUM || regnum == PS_REGNUM)
+    return builtin_type_int32;
+
+  if (regnum >= A0_REGNUM && regnum <= A0_REGNUM + 7)
+    return builtin_type_void_data_ptr;
+
+  return builtin_type_int32;
 }
 
 /* Function: m68k_register_name

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: m68k_register_virtual_type

[Andrew Cagney]

> |> Er, yes.  (And eventually long_double to builtin_type_m68881_ext but the
> |> last one might have fallout related to register conversions).
> 
> I have now committed this patch.

Thanks!

To answer your question.  The m68k shouldn't have any fallout because 
it's eliminated the CONVERT_FROM and CONVERT_TO methods (just never know 
until someone tests it :-).

On the other hand, a target like the MIPS, which uses the convert 
methods, could have problems.   Those functions try to cover up GDB's 
inability to manipulate floating point registers that had a non-standard 
type.  Fortunatly that inability has been fixed, just need to flush the 
old code.

Andrew



> Andreas.
> 
> 2002-11-23  Andreas Schwab  <schwab@suse.de>
> 
> 	* m68k-tdep.c (m68k_register_virtual_type): Use architecture
> 	invariant return values.
> 
> --- m68k-tdep.c.~1.28.~	2002-11-23 00:53:11.000000000 +0100
> +++ m68k-tdep.c	2002-11-23 00:53:32.000000000 +0100
> @@ -159,18 +159,19 @@ m68k_register_virtual_size (int regnum)
>  static struct type *
>  m68k_register_virtual_type (int regnum)
>  {
> -  if (regnum == E_FPI_REGNUM)
> -    return lookup_pointer_type (builtin_type_void);
> -  else if ((unsigned) regnum >= E_FPC_REGNUM)
> -    return builtin_type_int;
> -  else if ((unsigned) regnum >= FP0_REGNUM)
> -    return builtin_type_long_double;
> -  else if (regnum == PS_REGNUM)
> -    return builtin_type_int;
> -  else if ((unsigned) regnum >= A0_REGNUM)
> -    return lookup_pointer_type (builtin_type_void);
> -  else
> -    return builtin_type_int;
> +  if (regnum >= FP0_REGNUM && regnum <= FP0_REGNUM + 7)
> +    return builtin_type_m68881_ext;
> +