Looks good @JoeZiminski ! I reviewed this by commit, as suggested
@niksirbi how did things go in the end with the NIU test action, can I help with anything? I took the lazy way and revereted back to not using for now :'D
We kinda found a functional workaround for movement which allows us to keep using the NIU test action as is.
In pyproject.toml we configure tox-conda to install pytables in the test environment (for datashuttle that would be rclone).
The problem was that tox-conda could not find the conda executable on some OSes (macos-14 and windows), but setting up and activating conda before the tests seems to have taken care of that.
This is not now going to be possible with the different calls to pytest based on OS (SSH tests on linux, not otherwise). Instead we'll have to keep up with NIU best practices manually.
We kinda found a functional workaround for
movement
which allows us to keep using the NIU test action as is.Basically what we do is the following:
pyproject.toml
we configuretox-conda
to installpytables
in the test environment (fordatashuttle
that would berclone
).The problem was that
tox-conda
could not find the conda executable on some OSes (macos-14 and windows), but setting up and activating conda before the tests seems to have taken care of that.Maybe worth a try for
datashuttle
as well?Originally posted by @niksirbi in https://github.com/neuroinformatics-unit/datashuttle/pull/314#pullrequestreview-1961314209