build_spack sporadically fails due to timeout

[BenjaminRodenberg]

The test build_spack sporadically fails due to a timeout. First some background:

The test tries to build the spack package for the python bindings using the recipe from this repository. To reduce the runtime to a sane level I use the image benjaminrueth/ci-spack-pyprecice-deps-1804 where all the necessary dependencies are already built. However, sometimes the test fails to pick up the provided spack packages and begins building everything from scratch. To avoid this enormous waste of resources I decided to implement a 5 minutes timeout.

The reason for the failing test is therefore the following: Sometimes the test does not pick up the provided dependencies and fails due to timeout. Why? Can we fix this? I already realized that running the test (after failure) a second or even third time results in a successful run. This is good enough for development, but not really a satisfactory workflow.

[BenjaminRodenberg]

This issue might get fixed via #89

[BenjaminRodenberg]

I found a hint here. The base image built all dependencies on haswell, but the actual test runs on broadwell. Since the architecture is different, spack starts to rebuild everything.

[BenjaminRodenberg]

In this test the runner is haswell and therefore the test succeeds.