[RFC:avr] use regcache instead of read_register

[RFC:avr] use regcache instead of read_register

[Theodore A. Roth]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 487 bytes --]

Hi,

The ARI caught this and it seems like a trivial change. There is a bit of 
confusion though in the use of current_regcache since the ARI says this:

  current regcache 57 Replace current_regcache with explict parameter

What would that explicit parameter be?

If this patch is ok, it brings the ARI count for the AVR down to 0.

Ted Roth

2003-06-19  Theodore A. Roth  <troth@openavr.org>

	* avr-tdep.c (avr_read_pc): Use regcache instead of read_register.
	(avr_read_sp): Ditto.


[-- Attachment #2: Type: TEXT/PLAIN, Size: 1094 bytes --]

2003-06-19  Theodore A. Roth  <troth@openavr.org>

	* avr-tdep.c (avr_read_pc): Use regcache instead of read_register.
	(avr_read_sp): Ditto.

Index: avr-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/avr-tdep.c,v
retrieving revision 1.64
diff -u -r1.64 avr-tdep.c
--- avr-tdep.c	20 Jun 2003 05:53:42 -0000	1.64
+++ avr-tdep.c	20 Jun 2003 06:09:27 -0000
@@ -324,12 +324,12 @@
 avr_read_pc (ptid_t ptid)
 {
   ptid_t save_ptid;
-  CORE_ADDR pc;
+  ULONGEST pc;
   CORE_ADDR retval;
 
   save_ptid = inferior_ptid;
   inferior_ptid = ptid;
-  pc = (int) read_register (AVR_PC_REGNUM);
+  regcache_cooked_read_unsigned (current_regcache, AVR_PC_REGNUM, &pc);
   inferior_ptid = save_ptid;
   retval = avr_make_iaddr (pc);
   return retval;
@@ -349,7 +349,10 @@
 static CORE_ADDR
 avr_read_sp (void)
 {
-  return (avr_make_saddr (read_register (AVR_SP_REGNUM)));
+  ULONGEST sp;
+
+  regcache_cooked_read_unsigned (current_regcache, AVR_SP_REGNUM, &sp);
+  return (avr_make_saddr (sp));
 }
 
 static int

Re: [RFC:avr] use regcache instead of read_register

[Andrew Cagney]

> Hi,
> 
> The ARI caught this and it seems like a trivial change. There is a bit of 
> confusion though in the use of current_regcache since the ARI says this:
> 
>   current regcache 57 Replace current_regcache with explict parameter
> 
> What would that explicit parameter be?
> 
> If this patch is ok, it brings the ARI count for the AVR down to 0.

The patch is great.   Don't be too worried about current_regcache for 
the moment.

As for what parameter will be added to read_pc, I can give you several 
guesses:

- made redundant by unwind_pc(frame)
- replaced with something that takes the thread id and stop status

The underlying problem is decr pc after break.  Cleaning up that also 
involves read_pc, write_pc, and the update frame pc hack.

Andrew


> 2003-06-19  Theodore A. Roth  <troth@openavr.org>
> 
> 	* avr-tdep.c (avr_read_pc): Use regcache instead of read_register.
> 	(avr_read_sp): Ditto.
> 
> 
> 
> 
> 2003-06-19  Theodore A. Roth  <troth@openavr.org>
> 
> 	* avr-tdep.c (avr_read_pc): Use regcache instead of read_register.
> 	(avr_read_sp): Ditto.
> 
> Index: avr-tdep.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/avr-tdep.c,v
> retrieving revision 1.64
> diff -u -r1.64 avr-tdep.c
> --- avr-tdep.c	20 Jun 2003 05:53:42 -0000	1.64
> +++ avr-tdep.c	20 Jun 2003 06:09:27 -0000
> @@ -324,12 +324,12 @@
>  avr_read_pc (ptid_t ptid)
>  {
>    ptid_t save_ptid;
> -  CORE_ADDR pc;
> +  ULONGEST pc;
>    CORE_ADDR retval;
>  
>    save_ptid = inferior_ptid;
>    inferior_ptid = ptid;
> -  pc = (int) read_register (AVR_PC_REGNUM);
> +  regcache_cooked_read_unsigned (current_regcache, AVR_PC_REGNUM, &pc);
>    inferior_ptid = save_ptid;
>    retval = avr_make_iaddr (pc);
>    return retval;
> @@ -349,7 +349,10 @@
>  static CORE_ADDR
>  avr_read_sp (void)
>  {
> -  return (avr_make_saddr (read_register (AVR_SP_REGNUM)));
> +  ULONGEST sp;
> +
> +  regcache_cooked_read_unsigned (current_regcache, AVR_SP_REGNUM, &sp);
> +  return (avr_make_saddr (sp));
>  }
>  
>  static int

Re: [RFC:avr] use regcache instead of read_register

[Theodore A. Roth]

On Fri, 20 Jun 2003, Andrew Cagney wrote:

> > Hi,
> >
> > The ARI caught this and it seems like a trivial change. There is a bit of
> > confusion though in the use of current_regcache since the ARI says this:
> >
> >   current regcache 57 Replace current_regcache with explict parameter
> >
> > What would that explicit parameter be?
> >
> > If this patch is ok, it brings the ARI count for the AVR down to 0.
>
> The patch is great.   Don't be too worried about current_regcache for
> the moment.
>
> As for what parameter will be added to read_pc, I can give you several
> guesses:
>
> - made redundant by unwind_pc(frame)
> - replaced with something that takes the thread id and stop status
>
> The underlying problem is decr pc after break.  Cleaning up that also
> involves read_pc, write_pc, and the update frame pc hack.
>
> Andrew

Thanks.

Committed.

I guess a similar change to d10v_read_pc.

Just noticed that the d10v seems to use d10v_unwind_sp instead of
*_read_sp. Is it a direct replacement to replace *_read_sp with
*_unwind_sp? I didn't see that before I did the commit.

Ted Roth

>
>
> > 2003-06-19  Theodore A. Roth  <troth@openavr.org>
> >
> > 	* avr-tdep.c (avr_read_pc): Use regcache instead of read_register.
> > 	(avr_read_sp): Ditto.
> >
> >
> >
> >
> > 2003-06-19  Theodore A. Roth  <troth@openavr.org>
> >
> > 	* avr-tdep.c (avr_read_pc): Use regcache instead of read_register.
> > 	(avr_read_sp): Ditto.
> >
> > Index: avr-tdep.c
> > ===================================================================
> > RCS file: /cvs/src/src/gdb/avr-tdep.c,v
> > retrieving revision 1.64
> > diff -u -r1.64 avr-tdep.c
> > --- avr-tdep.c	20 Jun 2003 05:53:42 -0000	1.64
> > +++ avr-tdep.c	20 Jun 2003 06:09:27 -0000
> > @@ -324,12 +324,12 @@
> >  avr_read_pc (ptid_t ptid)
> >  {
> >    ptid_t save_ptid;
> > -  CORE_ADDR pc;
> > +  ULONGEST pc;
> >    CORE_ADDR retval;
> >
> >    save_ptid = inferior_ptid;
> >    inferior_ptid = ptid;
> > -  pc = (int) read_register (AVR_PC_REGNUM);
> > +  regcache_cooked_read_unsigned (current_regcache, AVR_PC_REGNUM, &pc);
> >    inferior_ptid = save_ptid;
> >    retval = avr_make_iaddr (pc);
> >    return retval;
> > @@ -349,7 +349,10 @@
> >  static CORE_ADDR
> >  avr_read_sp (void)
> >  {
> > -  return (avr_make_saddr (read_register (AVR_SP_REGNUM)));
> > +  ULONGEST sp;
> > +
> > +  regcache_cooked_read_unsigned (current_regcache, AVR_SP_REGNUM, &sp);
> > +  return (avr_make_saddr (sp));
> >  }
> >
> >  static int
>
>
>

Re: [RFC:avr] use regcache instead of read_register

[Andrew Cagney]

> I guess a similar change to d10v_read_pc.
> 
> Just noticed that the d10v seems to use d10v_unwind_sp instead of
> *_read_sp. Is it a direct replacement to replace *_read_sp with
> *_unwind_sp? I didn't see that before I did the commit.

There's currently a catch.

GDB doesn't have a consistent story over what ``info frame'' and "$sp" 
should display for the "sp".  Like the old "$fp", it can end up 
displaying inconsistent values.  I've a partial patch to fix it, need to 
go back though.  However, in the mean time, you can remove read_sp and 
see what happens.

Andrew