The recent 1.1.0 release broke two different users of pluggy (devpi and conda). In order to avoid such things in the future, we should improve our pre-release testing.
I went through pluggy's reverse dependencies and found a few large projects which use pluggy (at least these are the ones I recognized). Here is the list, with notes on how to test them (I haven't verified the steps yet):
- airflow
- conda
- datasette
- To test:
pytest -m venv venvvenv/bin/pip install -e .[test] https://github.com/pytest-dev/pluggy/archive/refs/heads/main.zipvenv/bin/pytest
- devpi
- hatch
- To test:
pytest -m venv venvvenv/bin/pip install -e . -e ./backend https://github.com/pytest-dev/pluggy/archive/refs/heads/main.zipvenv/bin/hatch run full
- pytest:
- To test:
- pytest continuously tests pluggy main in its CI
- tox
For now we can just suggest testing these manually in the RELEASING file, but in the future maybe we can automate it.
The recent 1.1.0 release broke two different users of pluggy (devpi and conda). In order to avoid such things in the future, we should improve our pre-release testing.
I went through pluggy's reverse dependencies and found a few large projects which use pluggy (at least these are the ones I recognized). Here is the list, with notes on how to test them (I haven't verified the steps yet):
pytest -m venv venvvenv/bin/pip install -e .[test] https://github.com/pytest-dev/pluggy/archive/refs/heads/main.zipvenv/bin/pytestpytest -m venv venvvenv/bin/pip install -e . -e ./backend https://github.com/pytest-dev/pluggy/archive/refs/heads/main.zipvenv/bin/hatch run fulltox -e py-pluggymainFor now we can just suggest testing these manually in the RELEASING file, but in the future maybe we can automate it.