On Nov 19, 2009, Richard Guenther <richard.guenther@gmail.com> wrote:

> On Thu, Nov 19, 2009 at 5:27 AM, Alexandre Oliva <aoliva@redhat.com> wrote:
>> On Nov 17, 2009, Richard Guenther <richard.guenther@gmail.com> wrote:
>> 
>>> Well, just adjust degenerate_phi_result to do instead of calling
>>> operand_equal_p
>> 
>>>   else if (TREE_CODE (arg) != TREE_CODE (val)
>>>              || (TREE_CODE (arg) == SSA_NAME
>>>                  && arg != val)
>>>              || !operand_equal_p (arg, val, 0))
>>>     break;
>> 
>>> that should fix it.
>> 
>> Good idea.

>> Here's what I'm testing now.

> Ok if it passes testing.

Thanks, but arg could be NULL, and then TREE_CODE (val) would crash.

The fixed patch below corrects the two regressions I got testing the
patch on x86_64-linux-gnu.  Save for new regressions, I'll check in the
revised version as soon as my current round of testing completes.