Closed trym-b closed 4 years ago
Tried this on my windows PC. It produces 3 errors, of which all seem reasonable:
FAILED test/test_camera.py::test_illegal_init - RuntimeError: Unsupported type for argument internal camera: this should fail, type: <class 'str'>.
FAILED test/test_point_cloud.py::test_illegal_init - ValueError: Unsupported type for argument internal_point_cloud. Got <class 'str'>, expected <class 'pybind11_builtins.pybind11_type'>
FAILED test/test_zivid_module_content.py::test_version - AssertionError: assert '0.9.1.1.7.0' == '1.0.0.1.8.1'
The first two failures "should fail". However, shouldn't a wanted failure be logged as success?
I tested this on master, but against python -m pip install zivid
. Why would zivid.__version__
be 0.9.1.1.7.0?
The reason it fails is that run_samples
assumes that test.py
is executed from continuous-integration/windows
. However, that's not the case for windows. For Linux --workdir /host/continuous-integration/linux
as argument for the Docker image. We could use the workingDirectory
argument for script
. However, @trym-b recommended that we remove --workdir /host/continuous-integration/linux
. That's a bigger job than simply making samples test work on windows. That can be solved by removing the depenency on current working directory in run_samples
:
- current_working_directory = Path(os.getcwd()).resolve()
sample = (
- current_working_directory
- / ".."
- / ".."
- / "samples"
- / "sample_{name}.py".format(name=name)
+ Path(__file__) / ".." / ".." / "samples" / "sample_{name}.py".format(name=name)
For some unknown reason test_samples.py fails when running on windows
This should be investigated