Closed ianhi closed 2 years ago
@martinRenou I consolidated a lot of the tests because we were spending a lot of time waiting for conda envs to install so i figured we could save on that by just having build
and test
steps. Is there a reason that we should instead keep the separate?
Is there a reason that we should instead keep the separate?
The idea behind this is to test the generated wheel while running the tests. Though I am not strongly against getting rid of it.
Is there a reason that we should instead keep the separate?
The idea behind this is to test the generated wheel while running the tests. Though I am not strongly against getting rid of it.
I think that that behavior was is maintained here - we still have a separate build
and test
step. I meant that the pytest tests and the visual regression tests were split into different steps which resulted in extra full installs
notes for myself:
Check out the matrix of jupyterlab dependencies. Part of the issue may be due to this:
- python: '3.6'
dist: 'ipympl*.tar.gz'
jlab_version: 3
- python: '3.9'
dist: 'ipympl*.whl'
jlab_version: 2.3.2
there's also this warning: LOG Run-time JupyterLab version (3.3.4) is different than testing framework is built for (3.0.11). This could cause issues in testing.
We need to update Galata quite a bit probably. I'll do this in this PR if it does not pass the tests.
The failing tests are the ones against Matplotlib's main branch, so we can ignore them for now and fix those later.
Please update galata references
Bot running in https://github.com/matplotlib/ipympl/actions/runs/2902499214
I removed the JupyterLab 2 CI test, as it's quite old anyway now and it reduces our build matrix.
As said above, the tests against latest Matplotlib are failing, let's ignore those for now.
Merging!
thank you martin!
Should fix #460