parallel test fails on 32-bit Fedora

[nalimilan]

I see this failure when running tests in my RPM package in 32-bits on a Fedora build machine.

cd test
make all
Warning: git information unavailable; versioning information limited
    �[32;1mJULIA�[0m �[37;1mtest/all�[0m
    From worker 2:       �[1m*�[0m �[31mlinalg1�[0m
    From worker 3:       �[1m*�[0m �[31mlinalg2�[0m
    From worker 4:       �[1m*�[0m �[31mlinalg3�[0m
    From worker 4:       �[1m*�[0m �[31mcore�[0m
    From worker 4:       �[1m*�[0m �[31mkeywordargs�[0m
    From worker 4:       �[1m*�[0m �[31mnumbers�[0m
    From worker 2:       �[1m*�[0m �[31mstrings�[0m
    From worker 2:       �[1m*�[0m �[31mcollections�[0m
    From worker 4:       �[1m*�[0m �[31mhashing�[0m
    From worker 2:       �[1m*�[0m �[31mremote�[0m
    From worker 3:       �[1m*�[0m �[31miobuffer�[0m
    From worker 2:       �[1m*�[0m �[31marrayops�[0m
    From worker 3:       �[1m*�[0m �[31msimdloop�[0m
    From worker 3:       �[1m*�[0m �[31mblas�[0m
    From worker 4:       �[1m*�[0m �[31mfft�[0m
    From worker 3:       �[1m*�[0m �[31mdsp�[0m
    From worker 3:       �[1m*�[0m �[31msparse�[0m
    From worker 4:       �[1m*�[0m �[31mbitarray�[0m
    From worker 3:       �[1m*�[0m �[31mrandom�[0m
    From worker 3:       �[1m*�[0m �[31mmath�[0m
    From worker 3:       �[1m*�[0m �[31mfunctional�[0m
    From worker 2:       �[1m*�[0m �[31mbigint�[0m
    From worker 3:       �[1m*�[0m �[31msorting�[0m
    From worker 2:       �[1m*�[0m �[31mstatistics�[0m
    From worker 3:       �[1m*�[0m �[31mspawn�[0m
    From worker 3:         �[34m[stdio passthrough ok]�[0m
    From worker 3:       �[1m*�[0m �[31mpriorityqueue�[0m
    From worker 2:       �[1m*�[0m �[31marpack�[0m
    From worker 3:       �[1m*�[0m �[31mfile�[0m
    From worker 2:       �[1m*�[0m �[31msuitesparse�[0m
    From worker 2:       �[1m*�[0m �[31mversion�[0m
    From worker 2:       �[1m*�[0m �[31mresolve�[0m
    From worker 3:       �[1m*�[0m �[31mpollfd�[0m
    From worker 3:       �[1m*�[0m �[31mmpfr�[0m
    From worker 4:       �[1m*�[0m �[31mbroadcast�[0m
    From worker 2:       �[1m*�[0m �[31mcomplex�[0m
    From worker 2:       �[1m*�[0m �[31msocket�[0m
    From worker 3:       �[1m*�[0m �[31mfloatapprox�[0m
    From worker 2:       �[1m*�[0m �[31mreaddlm�[0m
    From worker 3:       �[1m*�[0m �[31mregex�[0m
    From worker 3:       �[1m*�[0m �[31mfloat16�[0m
    From worker 4:       �[1m*�[0m �[31mcombinatorics�[0m
    From worker 3:       �[1m*�[0m �[31msysinfo�[0m
    From worker 3:       �[1m*�[0m �[31mrounding�[0m
    From worker 2:       �[1m*�[0m �[31mranges�[0m
    From worker 4:       �[1m*�[0m �[31mmod2pi�[0m
    From worker 3:       �[1m*�[0m �[31meuler�[0m
    From worker 4:       �[1m*�[0m �[31mshow�[0m
    From worker 3:       �[1m*�[0m �[31mlineedit�[0m
    From worker 4:       �[1m*�[0m �[31mreplcompletions�[0m
    From worker 3:       �[1m*�[0m �[31municode�[0m
    From worker 3:       �[1m*�[0m �[31mparallel�[0m
exception on 3: ERROR: test failed: et >= 1.0 && et <= 1.5
while loading parallel.jl, in expression starting on line 156
ERROR: test failed: et >= 1.0 && et <= 1.5
while loading parallel.jl, in expression starting on line 156
while loading /builddir/build/BUILD/julia-master/test/runtests.jl, in expression starting on line 51
WARNING: Forcibly interrupting busy workers
error in running finalizer: error in running finalizer: error in running finalizer: InterruptException()julia-debug: gc.c:303: run_finalizers: Assertion `ff != ((void*)1)' failed.
WARNING: Unable to terminate all workers
make: *** [all] Error 1

[amitmurthy]

Is the system under heavy load / thrashing? I have seen that particular error only when the system is low on memory and swapping a lot.

[nalimilan]

Hard to tell, as these are Fedora VMs. But it's clearly a possibility. The problem is, we're going to have to get this to work under these conditions...

[amitmurthy]

We could change tests where any sort of timing is involved to warnings, instead of bailing out on them... @StefanKarpinski @JeffBezanson opinions?

[JeffBezanson]

I think the test is too sensitive. Something taking 0.6 seconds instead of 0.5 is not really grounds for failing a test in a non-real-time system.

[amitmurthy]

The value of et should be just more than 1.0 seconds. The 0.5 seconds is the buffer since it is a non-real-time system. Think of sleep(1.0) actually sleeping for 1.6 seconds. We could increase the buffer, but it would just hide cases where a bug is causing unnecessary delays in notifications.