broune commented on issue #4367: [Tests] Dump duration of Pytests. URL: https://github.com/apache/incubator-tvm/pull/4367#issuecomment-555662599 tldr: how about having a run_pytets.sh file that takes care of how to run pytest? 1) I agree with Ramana that it is unfortunate to duplicate this into every invocation of pytest. 2) It is preferable to have it be very clear how to reproduce a command line run in CI and Docker in some other context and dependencies on environment variables make that harder. 3) Thinking ahead, we also want to run e.g. pytest-parallel, as the people on this thread have discussed before, which will also involve a parameter to pytest to trigger it to run N-ways parallel (providing the N) and ideally that will be done the same way. In this case we probably don't want to rebuild a different docker for each level of parallelism, which may vary by machine and other factors, so it's not ideal to handle pytest parameters via the Docker build from that perspective. I think we can satisfy 1-3 at the same time by having a run_pytest.sh script that is used to run pytest every time and can print out the actual command used. It can be changed without rebuilding the Dockers, it's in one place, so no duplication, it's easy to tell what it's doing and it gets us out of the current business (also before this changer) of encoding into each file how pytest is run. We can even have it print out a command line for how it was run, including any relevant environment variables like TVM_FFI and indeed PYTEST_ADDOPTS if they are non-empty. What do you think?
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
