Closed DJDavies2 closed 2 years ago
I don't think we can remove the timeout. It was there to prevent continuous integration to timeout at the default 1500 seconds, which must be happening sometimes. In my experience the 5s was always enough. What is the platform or compiler you're using? I am OK to increase this number. Perhaps 100 is too long still? What is the maximum variability you encountered?
I've done a series of runs. Often the times are around 15 seconds. Occasionally they are longer but perhaps that doesn't matter so much. I have push a change dropping the timeout to 20, would that be okay?
This is on a Cray with Intel and Gnu compilers. It also has the jobs running in parallel, maybe that is a factor? I don't know why it would be.
Is the solution in #21 okay?
Perfect! Thanks! Merged in develop.
I am getting transient timeout failures in fckit_test_configuration_fails on some platforms. Increasing the timeout to, say, 100 seems to work in that the failures go away. There is a big variation in runtime within those limits though.
Can the timeout be increased for this test? Or can the timeout be removed altogether?