Hi, Ken -
> I've reviewed these Frank, and cherry-picked 'em into my tree.
> qa/518 passes, but that does not prove much because I was not seeing any
> failures with this test on my QA machines.
Yeah, figured.
> [...] I wonder about the first one, as it seems killing -a pmie may
> produce collateral damage and indeed on the one system I tried this
> a pmie instance that has nothing to do with qa/518 was nuked. This
> does not seem right.
It is a bit confusing. Note that near the top of the 518 test case,
there was already:
# real QA test starts here
$sudo $signal -a pmie >/dev/null 2>&1
... so pmie instances not associated with the test are already on the
hit list.
> Do you have any additional information on the circumstances in which the
> first kill does not cause the pmie process launched by qa/518 to exit?
I'll try to trace it with something like systemtap. (Even with pmmgr
I encountered cases where a single SIGTERM sent to pmie was
blocked/ignored, so sudo is probably not a necessary component of the
problem.)
- FChE
|