On 2019-01-30 12:58, Brad Chamberlain wrote:
> Is it just that if
> something else is running intensively on the same node simultaneously,
> it may just take awhile to wake up? (I guess we don't have 'top' output
> in correctness testing to show whether this was the case for this
> configuration last night...?)
Yes, generally I would expect such slowness to be caused by
something else running on the same node. I would note that yesterday
there was considerable intermittent remote filesystem slowness, and if
that interfered in any way with the test system, it could have caused a
delay, so it might have been due to conditions on some other part of the
network.
Right, we don't have 'top' output for this test configuration.
> AFAICT, sleepTimeUnits only checks for undersleeping... I suppose a
> better way to make sure sleep wasn't taking way too long would be to
> make it into a performance test, in which case noise would be
> reasonable, but if a sleep() started multiplying its argument by 2 for
> some reason, we'd find out about it...? (though I realize that's
> starting to sound paranoid).
I think it's more likely that if something goes seriously wrong
with sleep, it would cause us to exceed our timeouts rather than just
multiply its argument by 2. Not that it couldn't happen.
David
> Thanks,
> -Brad
>
>
> On Wed, 30 Jan 2019, David Keaton wrote:
>
>> This is a test that sleeps for 5 seconds, and then checks to see
>> that it didn't sleep for more than 5.2 seconds. Elliot points out that
>> this is a bad idea, and that
>> test/library/standard/Time/sleepTimeUnits.chpl is a better timer test.
>>
>> Does anyone object to removing trivial/deitz/other/timer2? If I
>> hear no objections by tomorrow morning, I'll remove it.
>>
>> You don't have to give a reason for wanting to keep it. If
>> anybody wants it, I won't do anything.
>>
>> David
>>
>> On 2019-01-30 06:27, [email protected] wrote:
>>> === Summary ===================================================
>>> Hostname: chapcs10
>>> Revision: https://github.com/chapel-lang/chapel/commit/aba7f3c5d9
>>> Logfile: /data/sea/chapel/Nightly/day3-Wed-gasnet-everything.log
>>> Started: Wed Jan 30 02:04:44 2019
>>> Ended: Wed Jan 30 08:27:34 2019
>>> Tests run:
>>>
>>> View this job at:
>>>
>>> http://chapel-ci.us.cray.com/job/chapcs-correctness-test-gasnet-everything/928/
>>>
>>>
>>>
>>> 9534 Tests (+16), 796 Futures (-1)
>>> Tests run: 9533 Successes (+15), 1 Failures (+1)
>>> ===============================================================
>>>
>>> --- New Errors -------------------------------
>>> [Error matching program output for trivial/deitz/other/timer2]
>>>
>>> --- Resolved Errors --------------------------
>>>
>>> --- New Passing Future tests------------------
>>>
>>> --- Passing Future tests ---------------------
>>>
>>> --- New Passing Suppress tests------------------
>>>
>>> --- Passing Suppress tests ---------------------
>>>
>>> --- Unresolved Errors ------------------------
>>>
>>> --- New Failing Future tests -----------------
>>>
>>> ===============================================================
>>> Chapel Environment:
>>> CHPL_HOST_PLATFORM: linux64 *
>>> CHPL_HOST_COMPILER: gnu
>>> CHPL_HOST_ARCH: x86_64
>>> CHPL_TARGET_PLATFORM: linux64
>>> CHPL_TARGET_COMPILER: gnu
>>> CHPL_TARGET_ARCH: x86_64
>>> CHPL_TARGET_CPU: none *
>>> CHPL_LOCALE_MODEL: flat
>>> CHPL_COMM: gasnet *
>>> CHPL_COMM_SUBSTRATE: udp
>>> CHPL_GASNET_SEGMENT: everything *
>>> CHPL_TASKS: qthreads
>>> CHPL_LAUNCHER: amudprun
>>> CHPL_TIMERS: generic
>>> CHPL_UNWIND: none
>>> CHPL_MEM: jemalloc
>>> CHPL_ATOMICS: cstdlib
>>> CHPL_NETWORK_ATOMICS: none
>>> CHPL_GMP: gmp
>>> CHPL_HWLOC: hwloc
>>> CHPL_REGEXP: re2
>>> CHPL_LLVM: none
>>> CHPL_AUX_FILESYS: none
>>> CHPL_LIB_PIC: none
>>>
>>
>> _______________________________________________
>> Chapel-developers mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/chapel-developers
>>
_______________________________________________
Chapel-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/chapel-developers
