----- Original Message -----
> Hi guys,
>
> Firstly, thanks so much for that qa/src/multithread4.c fix
> Ken (the "fix ugly race" one) - that intermittent failure
> has plagued me for ages, and I'd never been able to quite
> put my finger on it. Thanks!
Hmm - I spoke too soon, and still see intermittent 449 failures.
$ ./check -q -l 449
449 0s ...
Passed all 1 tests
$ ./check -q -l 449
449 1s ...
Passed all 1 tests
$ ./check -q -l 449
449 1s ...
Passed all 1 tests
$ ./check -q -l 449
449 1s ...
Passed all 1 tests
$ ./check -q -l 449
449 0s ... - output mismatch (see 449.out.bad)
32c32
< pmStore: No permission to perform requested operation
---
> pmStore: Operation not supported for multi-threaded applications
Check local PMCD is still alive ...
PMDA probe: pminfo -h smash -f sample.milliseconds
PMDA probe: pminfo -h smash -f sampledso.milliseconds
PMDA probe: pminfo -h smash -f simple.numfetch
Failures: 449
Failed 1 of 1 tests
... which is the same issue I'm used to seeing on occasion.
> Dave, I have a consistent failure on 260 now ... .bad file
> vs .out shows:
>
> 324c324
> < r: 0.4
> ---
> > r: 0.50
>
> I've never seen this before, and its just started since the
> pmlogger/pmlc merge. For the life of me, I can't see how
> that could have affected this test though. It seems like we
> have someone sending an additional PDU to pmcd at some point.
> I'll continue to look into it next week, just thought I'd put
> it out there in case anyone else is seeing it and/or has any
> ideas.
In other strange sightings, test 260 spontaneously begun to
pass again (sometimes), and worse - the bug has multiplied!
Its evil offspring have jumped the fenceand test 259 has now
started to fail occasionally. Both are derived metrics tests
and, yeah, nothings changed in this area recently.
$ ./check -q -l 259
259 2s ... - output mismatch (see 259.out.bad)
147c147
< __dmpostfetch: [0] root node 511.0.2: numval=1 vset[0]: inst=-1 u=3
---
> __dmpostfetch: [0] root node 511.0.2: numval=1 vset[0]: inst=-1 u=5
161,162c161,162
< value 3
< 3
---
> value 5
> 5
Check local PMCD is still alive ...
PMDA probe: pminfo -h smash -f sample.milliseconds
PMDA probe: pminfo -h smash -f sampledso.milliseconds
PMDA probe: pminfo -h smash -f simple.numfetch
Failures: 259
Failed 1 of 1 tests
$ ./check -q -l 260
260 4s ...
Passed all 1 tests
$ ./check -q -l 259 260
259 2s ...
260 3s ... - output mismatch (see 260.out.bad)
348c348
< r: 0.50
---
> r: 0.57
Check local PMCD is still alive ...
PMDA probe: pminfo -h smash -f sample.milliseconds
PMDA probe: pminfo -h smash -f sampledso.milliseconds
PMDA probe: pminfo -h smash -f simple.numfetch
Failures: 260
Failed 1 of 2 tests
cheers.
--
Nathan
|