#176 shows that file distribution can be disrupted from dataset changes.
It would be good to have github action tests to run on all the datasets to ensure that the files in the main branch are publicly accessible and downloadable. This will avoid users running some tools that depend on templateflow to fail.
This is something we implemented for cneuromod https://github.com/courtois-neuromod/actions-template/blob/main/.github/workflows/deploy.yml
https://github.com/courtois-neuromod/actions-template/blob/main/tests/deploy/test_deploy.py
Every push to the main branch and PRs test file availability in public remote (among other tests, this is a bit custom).
The tests do not download the full files, but check that expected files are registered in certain special remotes and run git-annex fsck --fast to check their access (like a HTTP HEAD depending on the remote type).
WDYT?
#176 shows that file distribution can be disrupted from dataset changes.
It would be good to have github action tests to run on all the datasets to ensure that the files in the main branch are publicly accessible and downloadable. This will avoid users running some tools that depend on templateflow to fail.
This is something we implemented for cneuromod https://github.com/courtois-neuromod/actions-template/blob/main/.github/workflows/deploy.yml
https://github.com/courtois-neuromod/actions-template/blob/main/tests/deploy/test_deploy.py
Every push to the main branch and PRs test file availability in public remote (among other tests, this is a bit custom).
The tests do not download the full files, but check that expected files are registered in certain special remotes and run
git-annex fsck --fastto check their access (like a HTTP HEAD depending on the remote type).WDYT?