Fix test for convolution which fails from time to time
Closed this issue · 1 comments
Describe the bug
As it appeared coincidentally during the code review, one of the tests for the new convolution feature indeed fails from time to time in the CI pipeline. We should try to improve the criteria or test setup, to avoid occasionally failing tests.
See here for a failing example.
To Reproduce
Steps to reproduce the behavior:
- trigger any pipeline run
- watch the pipeline execute
- occasionally one of the test jobs fails (like here)
Expected behavior
The pipeline passes every time, until some errors are actually introduced.
Environment:
- CircleCI
The simple solution now seems to increase the Monte Carlo runs. This was enabled by introducing the slow marker for pytest, which excludes some expensive tests from the regular runs and allows for extensive nightly testing. We already increased the number of runs to 40000 in #224. This seems to be resolved. Thus we close this issue. In case thhe error occurs any time in the future, we will reopen this.