[PATCH] Fix linespec.c vs padding on 32bit targets

[PATCH] Fix linespec.c vs padding on 32bit targets

[Andrew Pinski]

[-- Attachment #1: Type: text/plain, Size: 428 bytes --]

Hi,
  The problem here is there is padding in address_entry so
iterative_hash_object on the full object cannot be done as the padding
would be some junk.  This causes lots of gdb failures on
mips64-linux-gnu.

OK?  Tested on mips64-linux-gnu.

Thanks,
Andrew Pinski

PS I don't have write access for gdb.


gdb/ChangeLog:
* linespec.c (hash_address_entry): Use iterative_hash_object on each
field rather than the struct itself.

[-- Attachment #2: fixhash.diff.txt --]
[-- Type: text/plain, Size: 621 bytes --]

? .elfread.c.swp
Index: linespec.c
===================================================================
RCS file: /cvs/src/src/gdb/linespec.c,v
retrieving revision 1.133
diff -u -p -r1.133 linespec.c
--- linespec.c	6 Dec 2011 19:57:47 -0000	1.133
+++ linespec.c	9 Dec 2011 03:49:07 -0000
@@ -253,8 +253,9 @@ static hashval_t
 hash_address_entry (const void *p)
 {
   const struct address_entry *aep = p;
-
-  return iterative_hash_object (*aep, 0);
+  hashval_t hash;
+  hash = iterative_hash_object (aep->pspace, 0);
+  return iterative_hash_object (aep->addr, hash);
 }
 
 /* An equality function for address_entry.  */

Re: [PATCH] Fix linespec.c vs padding on 32bit targets

[Tom Tromey]

>>>>> "Andrew" == Andrew Pinski <andrew.pinski@caviumnetworks.com> writes:

Andrew>   The problem here is there is padding in address_entry so
Andrew> iterative_hash_object on the full object cannot be done as the padding
Andrew> would be some junk.  This causes lots of gdb failures on
Andrew> mips64-linux-gnu.

Thanks Andrew.

Andrew> OK?  Tested on mips64-linux-gnu.

One tiny nit, see below.

Andrew> PS I don't have write access for gdb.

If you have write access to binutils then you do.
If you still don't, let me know and I will commit it.

Andrew> +  hashval_t hash;
Andrew> +  hash = iterative_hash_object (aep->pspace, 0);
Andrew> +  return iterative_hash_object (aep->addr, hash);

The gdb style puts a blank line between declarations and code.
Ok with that change.

Tom

Re: [PATCH] Fix linespec.c vs padding on 32bit targets

[Andrew Pinski]

On Fri, 2011-12-09 at 07:40 -0800, Tom Tromey wrote:
> >>>>> "Andrew" == Andrew Pinski <andrew.pinski@caviumnetworks.com> writes:
> 
> Andrew>   The problem here is there is padding in address_entry so
> Andrew> iterative_hash_object on the full object cannot be done as the padding
> Andrew> would be some junk.  This causes lots of gdb failures on
> Andrew> mips64-linux-gnu.
> 
> Thanks Andrew.
> 
> Andrew> OK?  Tested on mips64-linux-gnu.
> 
> One tiny nit, see below.
> 
> Andrew> PS I don't have write access for gdb.
> 
> If you have write access to binutils then you do.
> If you still don't, let me know and I will commit it.
> 
> Andrew> +  hashval_t hash;
> Andrew> +  hash = iterative_hash_object (aep->pspace, 0);
> Andrew> +  return iterative_hash_object (aep->addr, hash);
> 
> The gdb style puts a blank line between declarations and code.
> Ok with that change.

Thanks I committed the patch.

Thanks,
Andrew Pinski