Stefan Marsiske

hmmm, i don't quite see why the build is failing here, i don't think it is related to my changes.

> > is there a way to have pass-audit being used during a import? > > Yes, I can call pass-audit from pass-import like i searched in this repo for...

that is so awesome!

yeah, that's what i figured too, i thought maybe you do some random magic deeper in your mock/unit testing framework.

without my patches i cannot compile on alpine linux, so yeah i have these patches applied: https://github.com/aports-ugly/aports/tree/master/ugly/compute-runtime

i forgot to also note the modification that is mentioned in #277 supporting mmap without hugetlb, but i suspect that might be unrelated here.

yeah, i also tried to `DISABLE_` prefix that testname and see if the other tests succeed consistently, but no. i'll try some printf tracing.

so i put trace output in the constructor, destructor and `PageFaultManagerLinux::pageFaultHandlerWrapper()`. when everything goes right it gives this output: ``` set timeout to: 45 Iteration: 1 Iteration: 1. random_seed: installed...

> Is this a sporadic problem ? yeah it's non-deterministic. how non-deterministic? i just ran this: ``` sh fail=0; for a in $(seq 1000); do ./neo_page_fault_manager_tests || fail=$((fail+1)) ; done...

for reference: this is the test code that fails: ```c++ TEST(PageFaultManagerLinuxTest, whenPageFaultIsRaisedThenHandlerIsInvoked) { auto pageFaultManager = std::make_unique(); EXPECT_FALSE(pageFaultManager->handlerInvoked); std::raise(SIGSEGV); EXPECT_TRUE(pageFaultManager->handlerInvoked); } ``` source: https://github.com/intel/compute-runtime/blob/03ce6681a052a6009ee10649617395a1e83138e5/shared/test/unit_test/page_fault_manager/linux/cpu_page_fault_manager_linux_tests.cpp#L26 it fails in the last line...