Intermittent core dump in native pair hmm about gkl HOT 5 CLOSED

@cmnbroad Thanks for reporting. I tagged @SnehalA and @Kmannth on the GATK issue.

from gkl.

@cmnbroad

The team did a Java 11 test with GKL Master/v0.8.8.

compiled GATK with Java 11 with the Master GKL from our repository and then ran the following tests over the weekend:

• IntelInflaterDeflaterIntegrationTest
• HaplotypeUnitTest
• SmithWatermanIntelAlignerUnitTest
• VectorPairHMMUnitTest
  These 4 tests have executed upwards of 20,000 times and not a single crash was encountered. It's worth noting that the last test (VectorPairHMMUnitTest) was meant to be disabled but didn't actually skip the test due to an incorrect Java version check.

We have been unable to reproduce this with latest code Master/v0.8.8.

Can you think of any other thing we should try ?

from gkl.

@kmmanth I'm not sure what you mean by:

It's worth noting that the last test (VectorPairHMMUnitTest) was meant to be disabled but didn't actually skip the test due to an incorrect Java version check.

Are you saying the code that disables the test is incorrect ? Can you provide more detail ?

from gkl.

@cmnbroad
In the current code there is check to skip VectorPairHMMUnitTest with a java version check. On our Centos 7.x dev system the way Java was being check did not work so it ran anyway. I don't know exactly why the Java version check might work on your system but not ours but it is something we noticed.

from gkl.

We believe this is fixed by the latest version. Please reopen if you hit it again.

from gkl.