Re: [PATCH] Fix agent code generate bug of ref

[Hui Zhu <hui_zhu@mentor.com>]

On 03/12/13 22:16, Tom Tromey wrote:
>>>>>> "Hui" == Hui Zhu <teawater@gmail.com> writes:
>
> Hui> Sorry for my misunderstand in your mail.  Accord to discussion with
> Hui> Yao in IRC.  I merge 2 patches together.
> Hui> And I found that gdb_assert of access_memory.  It should use nbytes.
>
> Thanks for doing this.
>
> Hui> 2013-03-12  Yao Qi  <yao@codesourcery.com>
> Hui> 	    Hui Zhu  <hui_zhu@mentor.com>
> Hui> 	* dwarf2loc.c (access_memory): Change nbits to nbytes in gdb_assert.
> Hui> 	(dwarf2_compile_expr_to_ax): Call access_memory in DW_OP_deref and
> Hui> 	DW_OP_deref_size.
>
> Hui> +  gdb_assert (nbytes > 0 && nbytes <= sizeof (LONGEST));
>
> Hui> -	      default:
> Hui> -		/* Note that get_DW_OP_name will never return
> Hui> -		   NULL here.  */
> Hui> -		error (_("Unsupported size %d in %s"),
> Hui> -		       size, get_DW_OP_name (op));
> Hui> -	      }
>
> I think we need a sanity check before calling access_memory.
> Otherwise, bad DWARF will be able to crash gdb.
> The patch is ok with that change.
>
> Tom
>

According to the discussion with Tom in IRC.
I add a check before call access_memory.
+	    if (size != 1 && size != 2 && size != 4 && size != 8)
+	      error (_("Refn doesn't support size %d"),
+		     size * TARGET_CHAR_BIT);

Checked in http://sourceware.org/ml/gdb-cvs/2013-03/msg00102.html

Thanks,
Hui