[RFA] Update D10V target to use new Harvardized type casts.

[RFA] Update D10V target to use new Harvardized type casts.

[Michael Snyder]

This one-line change lets the D10V back-end recognize the new
Harvardized type casts eg. "(@code int *)", so that a pointer 
can be forced into the address space of choice.

The exact same change can be used for (most) other Harvard back-ends.

2001-11-14  Michael Snyder  <msnyder@redhat.com>

	* d10v-tdep.c (d10v_pointer_to_address): Use new type flag
	TYPE_FLAG_CODE_SPACE to recognize a pointer that has been cast
	into the instruction address space.

Index: d10v-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/d10v-tdep.c,v
retrieving revision 1.25
diff -p -r1.25 d10v-tdep.c
*** d10v-tdep.c	2001/10/15 18:18:29	1.25
--- d10v-tdep.c	2001/11/14 20:56:37
*************** d10v_pointer_to_address (struct type *ty
*** 413,419 ****
  
    /* Is it a code address?  */
    if (TYPE_CODE (TYPE_TARGET_TYPE (type)) == TYPE_CODE_FUNC
!       || TYPE_CODE (TYPE_TARGET_TYPE (type)) == TYPE_CODE_METHOD)
      return d10v_make_iaddr (addr);
    else
      return d10v_make_daddr (addr);
--- 413,420 ----
  
    /* Is it a code address?  */
    if (TYPE_CODE (TYPE_TARGET_TYPE (type)) == TYPE_CODE_FUNC
!       || TYPE_CODE (TYPE_TARGET_TYPE (type)) == TYPE_CODE_METHOD ||
!       (TYPE_FLAGS (TYPE_TARGET_TYPE (type)) & TYPE_FLAG_CODE_SPACE) != 0)
      return d10v_make_iaddr (addr);
    else
      return d10v_make_daddr (addr);

Re: [RFA] Update D10V target to use new Harvardized type casts.

[Andrew Cagney]

> This one-line change lets the D10V back-end recognize the new
> Harvardized type casts eg. "(@code int *)", so that a pointer 
> can be forced into the address space of choice.
> 
> The exact same change can be used for (most) other Harvard back-ends.

(I quit as d10v maintainer :-)

Yes, looks ok.

One thought, though, should that test:

	... == FUNC || ... == METHOD || ... == CODE_SPACE

be abstracted out?  Every harvard target I've looked at has had that 
same test.

	Andrew


> 2001-11-14  Michael Snyder  <msnyder@redhat.com>
> 
> * d10v-tdep.c (d10v_pointer_to_address): Use new type flag
> 	TYPE_FLAG_CODE_SPACE to recognize a pointer that has been cast
> 	into the instruction address space.
> 
> Index: d10v-tdep.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/d10v-tdep.c,v
> retrieving revision 1.25
> diff -p -r1.25 d10v-tdep.c
> *** d10v-tdep.c	2001/10/15 18:18:29	1.25
> --- d10v-tdep.c	2001/11/14 20:56:37
> *************** d10v_pointer_to_address (struct type *ty
> *** 413,419 ****
>   
>     /* Is it a code address?  */
>     if (TYPE_CODE (TYPE_TARGET_TYPE (type)) == TYPE_CODE_FUNC
> !       || TYPE_CODE (TYPE_TARGET_TYPE (type)) == TYPE_CODE_METHOD)
>       return d10v_make_iaddr (addr);
>     else
>       return d10v_make_daddr (addr);
> --- 413,420 ----
>   
>     /* Is it a code address?  */
>     if (TYPE_CODE (TYPE_TARGET_TYPE (type)) == TYPE_CODE_FUNC
> !       || TYPE_CODE (TYPE_TARGET_TYPE (type)) == TYPE_CODE_METHOD ||
> !       (TYPE_FLAGS (TYPE_TARGET_TYPE (type)) & TYPE_FLAG_CODE_SPACE) != 0)
>       return d10v_make_iaddr (addr);
>     else
>       return d10v_make_daddr (addr);
> 
> 

Re: [RFA] Update D10V target to use new Harvardized type casts.

[Michael Snyder]

Andrew Cagney wrote:
> 
> > This one-line change lets the D10V back-end recognize the new
> > Harvardized type casts eg. "(@code int *)", so that a pointer
> > can be forced into the address space of choice.
> >
> > The exact same change can be used for (most) other Harvard back-ends.
> 
> (I quit as d10v maintainer :-)
> 
> Yes, looks ok.

Committed.

> One thought, though, should that test:
> 
>         ... == FUNC || ... == METHOD || ... == CODE_SPACE
> 
> be abstracted out?  Every harvard target I've looked at has had that
> same test.

Possible.  Although it's only a few lines, so far they  have shaken out
as pretty generically useful.  Can't think where they would go though...

Re: [RFA] Update D10V target to use new Harvardized type casts.

[Andrew Cagney]

>> One thought, though, should that test:
>> 
>> ... == FUNC || ... == METHOD || ... == CODE_SPACE
>> 
>> be abstracted out?  Every harvard target I've looked at has had that
>> same test.
> 
> 
> Possible.  Although it's only a few lines, so far they  have shaken out
> as pretty generically useful.  Can't think where they would go though...

I'd guess a simple predicate function in gdbtypes.h + gdbtypes.c.  The 
reason for suggesting it is that I've been tweeking random targets so 
that they get the lines right.

Andrew