Re: [Patch, trivial] Fix a potential bug of cost_classes_eq in ira-costs.c

[Vladimir Makarov <vmakarov at redhat dot com>]

On 12/30/2013, 2:27 AM, Yangfei (Felix) wrote:
> Add one entry to ChangeLog:
>
> Index: gcc/ChangeLog
> ===================================================================
> --- gcc/ChangeLog       (revision 206236)
> +++ gcc/ChangeLog       (working copy)
> @@ -1,3 +1,8 @@
> +2013-12-30  Felix Yang  <felix.yang@huawei.com>
> +
> +       * gcc/ira-costs.c (cost_classes_hasher::equal): memcmp return 0 if no
> +       difference exists for HV1 and HV2.
> +
> Index: gcc/ira-costs.c
> ===================================================================
> --- gcc/ira-costs.c	(revision 206236)
> +++ gcc/ira-costs.c	(working copy)
> @@ -155,7 +155,7 @@ inline bool
>   cost_classes_hasher::equal (const value_type *hv1, const compare_type *hv2)
>   {
>     return hv1->num == hv2->num && memcmp (hv1->classes, hv2->classes,
> -					 sizeof (enum reg_class) * hv1->num);
> +					 sizeof (enum reg_class) * hv1->num) == 0;
>   }
>
>   /* Delete cost classes info V from the hash table.  */

Thanks for finding this typo and pointing this out.  The probability is 
very small that this typo results in different code generation (as a 
hash is used first for accessing hash table) and is zero for wrong code 
generation.  But it results in wasting GCC memory by duplication of the 
same cover class set in the hash table.

As for the patch.  The changelog entry should have ira-costs.c not 
gcc/ira-costs.c.  Another thing, the line should have less 80 chars.

I see that you have no write access to GCC repository.  I'll commit the 
modified version of your patch by myself.

Thanks again, Felix.