Re: powerpc remote target registers

[Andrew Cagney <cagney@gnu.org>]

[-- Attachment #1: Type: text/plain, Size: 2268 bytes --]

> Andrew Cagney wrote:
> 
> But then the registers aren't marked as cached at all, so they're now requested from the target each time you do "info all-registers", even though they come up with 0s. Should I pretend the registers not supplied by the target were 0, or should I mark them as unavailable (i.e. the same as what having an "x" does) so at least it's consistent?
> 
> 
> Ah, they should be supplied but with a value of zero.  The protocol (for  historic reasons) specifies that a short G packet should have the missing entries treated as zero (like you intended).
> 
> Good, in which case the attached patch (against 6.0) should do it. Mostly indent changes, boringly enough.
> 
> 
> 2003-12-04  Jonathan Larmour  <jifl@eCosCentric.com>
> 
>     * remote.c (remote_fetch_registers): If target doesn't supply
>     registers, set them to zero.
> 
> Thanks,

Try the attached, its basicly the same but with a few not very obvious 
tweaks: supply_register is actually deprecated (but you couldn't tell 
:-); ->offset is really only valid when ->in_g_packet; avoids an 
assuption about the total size of the buffer and the behavior of get packet.

I think I got the logic right.

Andrew

(PS: paperwork sent)

> --- remote.c.old	2003-12-02 03:05:46.000000000 +0000
> +++ remote.c	2003-12-04 07:19:38.000000000 +0000
> @@ -3498,19 +3498,31 @@ remote_fetch_registers (int regnum)
>  	warning ("Remote reply is too short: %s", buf);
>      }
>  
>   supply_them:
>    {
> -    int i;
> +    int i, end_targ_regs=0;
>      for (i = 0; i < NUM_REGS + NUM_PSEUDO_REGS; i++)
>        {
>  	struct packet_reg *r = &rs->regs[i];
> +
> +	if (buf[r->offset * 2] == 0)
> +	  end_targ_regs = 1;  /* end of registers supplied by target */
>  	if (r->in_g_packet)
>  	  {
> -	    supply_register (r->regnum, regs + r->offset);
> -	    if (buf[r->offset * 2] == 'x')
> -	      set_register_cached (i, -1);
> +	    if (end_targ_regs)
> +	      {
> +		/* If the target hasn't sent enough registers, set
> +		   the remainder to 0. */
> +		supply_register (r->regnum, 0);
> +	      }
> +	    else
> +	      {
> +		supply_register (r->regnum, regs + r->offset);
> +		if (buf[r->offset * 2] == 'x')
> +		  set_register_cached (i, -1);
> +	      }
>  	  }
>        }
>    }
>  }
>  



[-- Attachment #2: diffs --]
[-- Type: text/plain, Size: 1455 bytes --]

2003-12-04  Andrew Cagney  <cagney@redhat.com>

	* remote.c (remote_fetch_registers): For short packets, explicitly
	supply a zero value.  Use regcache_raw_supply.  Fix suggested by
	Jonathan Larmour.

Index: remote.c
===================================================================
RCS file: /cvs/src/src/gdb/remote.c,v
retrieving revision 1.122
diff -u -r1.122 remote.c
--- remote.c	10 Nov 2003 21:20:44 -0000	1.122
+++ remote.c	4 Dec 2003 15:05:49 -0000
@@ -3558,9 +3558,23 @@
 	struct packet_reg *r = &rs->regs[i];
 	if (r->in_g_packet)
 	  {
-	    supply_register (r->regnum, regs + r->offset);
-	    if (buf[r->offset * 2] == 'x')
-	      set_register_cached (i, -1);
+	    if (r->offset * 2 >= strlen (buf))
+	      /* A short packet that didn't include the register's
+                 value, this implies that the register is zero (and
+                 not that the register is unavailable).  Supply that
+                 zero value.  */
+	      regcache_raw_supply (current_regcache, r->regnum, NULL);
+	    else if (buf[r->offset * 2] == 'x')
+	      {
+		gdb_assert (r->offset * 2 < strlen (buf));
+		/* The register isn't available, mark it as such (at
+                   the same time setting the value to zero).  */
+		regcache_raw_supply (current_regcache, r->regnum, NULL);
+		set_register_cached (i, -1);
+	      }
+	    else
+	      regcache_raw_supply (current_regcache, r->regnum,
+				   regs + r->offset);
 	  }
       }
   }