From: Jin, Gordon <gordon.jin@in...> - 2005-07-12 00:55:59
Roman Kagan wrote:
> On Fri, Jul 08, 2005 at 11:10:57AM +0800, Gordon Jin wrote:
>> Running nptl01 takes quite long time (>10min on my machine).
>> It was not a big issue other than lasting time.
>> But with recent change v1.5 (happened on Apr 27), the test will
>> become FAIL if not completed within a hard limit (MAXTIME=3D300sec).
>> We have 2 choices to resolve:
>> 1) increase MAXTIME to a large enough value.
>> 2) decrease execution time.
>> I think 2) is more reasonable.
>> The attached patch reduces iterations from 1000000 to 100000 to
>> decrease execution time to about 1-2 min.
> Isn't it defeating the whole purpose of this testcase, which is to
> increase the probability of triggering the subtle bug by means of a
> big enough number of iterations? =20
But I can't see the substantial difference between 10^6 and 10^5...
Lasting for too long time does hurt the test: not only extending the
for whole test suite for 10min+, but also getting nptl01 fail.=20
And reducing to 1/10 time looks fine to me.
I'm interested to know how long nptl01 takes on anyone's machine?
I believe the failure not limited on my machine.