[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

hypre tests failing irreproducibly, is openmpi3 behaving well?



I tried to activate debci (autopkgtest) for hypre 2.14.0-4, but it's proving frustrating. Many arches (32 bit in particular but also mpis64el) fail to pass build-time tests. But even amd64 is failing debci tests, https://ci.debian.net/packages/h/hypre/unstable/amd64/

The frustrating thing is that the test failures are not reproducible. debci times out on a different test each time. Strange also that the debci tests (for hypre 2.14.0-4) succeeded on testing (15 min to complete), it's on unstable that they're failing.

A manual run of "autopkgtest -B -- null" succeeds on my own computer, and on the amd64 Debian project machine. That said, I do get occasional build-time time-outs when building the package on my own computer.

Upstream recently released hypre 2.15.0 so there's the off-chance they've fixed a bug that might help. I was hoping to have debci working smoothly with hypre before the transition, however.

The pattern seems similar to the openmpi RC bugs #905418, #907267, #907407, where tests are failing for other packages. Alastair, do we have any clue what's going on here? Any new OMPI_MCA_* environment variable we should consider activating? Or should I judge the hypre tests to be too "heavy" (unreliable) for automated testing?

Drew


Reply to: