Closed duylethanh closed 3 years ago
PR updated. I made the requested changes. ~, while doing some refactoring to the new code. I also removed unused import
statements.~
Are there any other simple but useful tests that we could add? Maybe running experiments via the Python API (= a fork launcher)? I guess we would have to copy the experiments directory to a temporary location for that to work.
I have added some cli tests via subprocess.check_call()
. As we only verify the return code with this method we might run into similar problems like the following;
simex e launch
will launch the experiments or skip experiments if the instance is not available. Both actions will not cause an error and the program ends with return code 0
.
This means, that we might want to parse the print()
outputs of a subprocess and compare it to the expected output, to avoid different outcomes with the same return code.
edit: I removed the unrelated commits to keep this PR clean. I will make a separate PR for the minor fixes.
This PR contains the following:
master
branchpytest
) forutil.validate_setup_file()
,simex e launch --launch-through=fork
andsimex e purge --all -f
pytest6.1.2
toinstall_requires
insetup.py
~The workflow currently uses the cross product of
[ubuntu-latest, macos-latest]
and[python3.6, python3.7]
to run the unit tests The following screenshot is from a test pull request on my fork https://github.com/duylethanh/simexpal/pull/4:Some things to consider:
ubuntu-latest
resolves to Ubuntu 18.04, although Ubuntu 20.04 is available as previewmacos-latest
resolves to macOS 10.15 (Catalina), although macOS 11.0 (BigSur) is available as previewpytest
, we might wanna relax that condition