autotest: refactor so that test steps can be run independently #4407
Assignees
Labels
Comments
4 tasks
|
Some lists below of failures I observe when running in random order - test files with failures, and a longer list of individual failures within those files. It's possible (probable) that neither is comprehensive, as the order was still close enough to "normal" for the tests to pass. The random seed can be altered with #4421 fixes the single example I found where changing the test order causes a segmentation fault and crashes the test runner. I also found that when running with random order that Pytest does not exit at the end. On giving a keyboard interrupt, the traceback shows it was stuck trying to join a thread, but gave no indication of where that thread was spawned from. Files with failuresTest failures |
|
@rouault after making the randomization suggestion, I began using the pytest-randomly plugin with rasterio's tests on Travis CI and on my laptop. Since then, I've been seeing sporadic proj_create failures with PROJ versions 7.2.1, 8.0.1, and 8.1.1 rasterio/rasterio#2282 (comment). Perhaps this is due to a flaw in rasterio or its tests, but perhaps it is a signal of something else. |
rouault
added
the
code enhancements
Merged
4 tasks
3 tasks
3 tasks
3 tasks
3 tasks
3 tasks
3 tasks
3 tasks
|
This has been completed through the PRs referenced above. |
|
Epic effort @dbaston ! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Currently a number of test files have interdependent steps. It would be good to make them independent. Using
pytest --random-order (https://pypi.org/project/pytest-random-order/)can be for example used to check thatThe text was updated successfully, but these errors were encountered: