Closed hugovk closed 9 months ago
I'm of the opinion we test
I recommend that what we test == what we support. So support Sphinx 5-7, and Python 3.9-3.12, dropping Sphinx 4 and Python 3.8?
I think it's ok to keep everything in. The only issue would be too many macos runners delaying allwpilib builds but this repo doesn't get daily activity so i don't think it's a big deal.
Well, it's also possible support. I agree we shouldn't support what we don't use. And since really this project was made for frc-docs, we should closely align to the update timelines we do there (major dependency/python updates once a year).
I think I agree with @hugovk and it doesn't really hurt us. Users on older Sphinx versions can pin to a lower opengraph version. It's not like we're booming with momentum.
Sphinx 7.0.0 was released on 2023-04-30 and we're now up to 7.2.6.
Let's test it.
However, combined with PR #113, that means we have 6 Python versions x 5 Sphinx versions x 3 operating systems = 90 jobs in the matrix!
I suggest we drop support for Sphinx 4 to bring it down to 6x4x3 = 72 jobs.
Alternatively, we could run the full matrix for Ubuntu, but selectively run a subset of versions for other operating systems. For example, for macOS and Windows, only test the lowest/highest supported Sphinx (4 and 7), and/or Python (3.8 and 3.12).