[PATCH] hppa: Optimize atomic_compare_and_exchange_val_acq

Carlos O'Donell carlos@redhat.com
Tue Sep 27 17:17:00 GMT 2016


On 09/23/2016 03:13 AM, Florian Weimer wrote:
> * John David Anglin:
> 
>>>> The change was intended as an optimization but tst-stack4 now passes.
>>>
>>> This is a red flag for this patch.
>>
>> The patch has had a significant amount of testing and I'm sure there
>> are no functional issues.  It is important to simplify
>> when possible.
>>
>> Why red flag an improvement in test results?
> 
> I share Carlos' concern.  If this patch appears to fix tst-stack4,
> then it has unforeseen consequences.  We just don't understand the
> full impact of this change.

Exactly.

Before and after the patch the regression testsuite should give
roughly the same results (modulo unreliable tests).
 
I agree that operational hours of testing are important, but it's
not everything one has to consider.

Is the FAIL->PASS for tst-stack4 reliably tied to the addition of
this patch?

-- 
Cheers,
Carlos.



More information about the Libc-alpha mailing list