[PATCH/RFA] builtin-regs buglet

[PATCH/RFA] builtin-regs buglet

[Richard Earnshaw]

[-- Attachment #1: Type: text/plain, Size: 304 bytes --]


I think this probably come under obvious, but lets see what others think.

With my new ARM pseudo-registers code, I'm getting an internal error if I 
type

(gdb) p $fp

Why?  Well there's some missing brackets...

	* builtin-regs.c (value_of_builtin_reg): Correctly calculate the
	builtin reg number.



[-- Attachment #2: gdb-builtin.patch --]
[-- Type: text/x-patch , Size: 839 bytes --]

Index: builtin-regs.c
===================================================================
RCS file: /cvs/src/src/gdb/builtin-regs.c,v
retrieving revision 1.1
diff -p -r1.1 builtin-regs.c
*** builtin-regs.c	9 Apr 2002 03:06:13 -0000	1.1
--- builtin-regs.c	13 May 2002 15:59:50 -0000
*************** builtin_reg_map_name_to_regnum (const ch
*** 71,77 ****
  struct value *
  value_of_builtin_reg (int regnum, struct frame_info *frame)
  {
!   int reg = regnum - NUM_REGS + NUM_PSEUDO_REGS;
    gdb_assert (reg >= 0 && reg < nr_builtin_regs);
    return builtin_regs[reg].value (frame);
  }
--- 71,77 ----
  struct value *
  value_of_builtin_reg (int regnum, struct frame_info *frame)
  {
!   int reg = regnum - (NUM_REGS + NUM_PSEUDO_REGS);
    gdb_assert (reg >= 0 && reg < nr_builtin_regs);
    return builtin_regs[reg].value (frame);
  }

Re: [PATCH/RFA] builtin-regs buglet

[Andrew Cagney]

> 	* builtin-regs.c (value_of_builtin_reg): Correctly calculate the
> 	builtin reg number.

Oops!  Yes.

Want to live dangerously?  Try setting fp_regnum to -1.  It should 
[finally] fix the $fp (regname) vs $fp (info registers) problem.

Andrew

Re: [PATCH/RFA] builtin-regs buglet

[Richard Earnshaw]

> > 	* builtin-regs.c (value_of_builtin_reg): Correctly calculate the
> > 	builtin reg number.
> 
> Oops!  Yes.

Installed.

> 
> Want to live dangerously?  Try setting fp_regnum to -1.  It should 
> [finally] fix the $fp (regname) vs $fp (info registers) problem.

Hmm, one step at a time... My current patch to for the arm is already 
looking somewhat jumbo like.  Unfortunately, I don't really see how I can 
break it down, since the changes are all fairly fundamental.

R.

Re: [PATCH/RFA] builtin-regs buglet

[Andrew Cagney]

> 	* builtin-regs.c (value_of_builtin_reg): Correctly calculate the
>> > 	builtin reg number.
> 
>> 
>> Oops!  Yes.
> 
> 
> Installed.
> 
> 
>> 
>> Want to live dangerously?  Try setting fp_regnum to -1.  It should 
>> [finally] fix the $fp (regname) vs $fp (info registers) problem.
> 
> 
> Hmm, one step at a time... My current patch to for the arm is already 
> looking somewhat jumbo like.  Unfortunately, I don't really see how I can 
> break it down, since the changes are all fairly fundamental.

Try cherry [mango] picking a little.  Stuff like the above, also no 
longer defining register_{raw,virtual}_size.

However, yes you're heading for something of a rewrite.

Andrew

Re: [PATCH/RFA] builtin-regs buglet

[Richard Earnshaw]

ac131313@cygnus.com said:
> However, yes you're heading for something of a rewrite. 

You said it...

The size of this, plus some of the attendant changes elsewhere, makes me 
wonder whether we should create a separate branch -- this could be quite 
destabilizing, particularly for some of the target connections that use 
the revised ARM code.

Thoughts?

R.

Re: [PATCH/RFA] builtin-regs buglet

[Andrew Cagney]

> 
> You said it...
> 
> The size of this, plus some of the attendant changes elsewhere, makes me 
> wonder whether we should create a separate branch -- this could be quite 
> destabilizing, particularly for some of the target connections that use 
> the revised ARM code.

Yes.  Could use the regbuf branch or create your own.  I can test 
against arm-sim and I could do with an extra target to test against.

As for the change, I suspect there are two steps: make everything a 
pseudo (probably the bit to have on the branch?); then add lots more to 
the raw/pseudo registers.

thoughts?
Andrew

Re: [PATCH/RFA] builtin-regs buglet

[Richard Earnshaw]

> > 
> > You said it...
> > 
> > The size of this, plus some of the attendant changes elsewhere, makes me 
> > wonder whether we should create a separate branch -- this could be quite 
> > destabilizing, particularly for some of the target connections that use 
> > the revised ARM code.
> 
> Yes.  Could use the regbuf branch or create your own.  I can test 
> against arm-sim and I could do with an extra target to test against.
> 
> As for the change, I suspect there are two steps: make everything a 
> pseudo (probably the bit to have on the branch?); then add lots more to 
> the raw/pseudo registers.

Putting it in the regbuf branch might be best, but it'll need the 
REGISTER_SIM_REGNO patch if I'm to get the sim target working.

R.