[tuples] [patches] small bug fix

[David Daney]

David Daney wrote:
> Diego Novillo wrote:
>   
>> Actually, I have a slight preference for:
>>
>>
>> Index: tree-phinodes.c
>> ===================================================================
>> --- tree-phinodes.c     (revision 132865)
>> +++ tree-phinodes.c     (working copy)
>> @@ -146,6 +146,8 @@ allocate_phi_node (size_t len)
>>      {
>>        free_phinode_count--;
>>        phi = VEC_pop (gimple, free_phinodes[bucket]);
>> +      if (VEC_empty (free_phinodes[bucket]))
>> +       VEC_free (free_phinodes[bucket]);
>>  #ifdef GATHER_STATISTICS
>>        phi_nodes_reused++;
>>  #endif
>>
>>
>> Does that fix your regression?
>>
>>   
>>     
> That doesn't compile for me.
>
> I am testing the attached.
>
>
>   
> ------------------------------------------------------------------------
>
> Index: tree-phinodes.c
> ===================================================================
> --- tree-phinodes.c	(revision 132948)
> +++ tree-phinodes.c	(working copy)
> @@ -146,6 +146,8 @@ allocate_phi_node (size_t len)
>      {
>        free_phinode_count--;
>        phi = VEC_pop (gimple, free_phinodes[bucket]);
> +      if (VEC_empty (gimple, free_phinodes[bucket]))
> +	VEC_free (gimple, gc, free_phinodes[bucket]);
>  #ifdef GATHER_STATISTICS
>        phi_nodes_reused++;
>  #endif
>   
And now I can confirm that this lets me build on mipsel-linux, where as
previously I got an ICE building libiberty.

David Daney