Should we pin versions here? about robotpy-meta HOT 7 CLOSED

I think best practice for stability would be pinning versions, or bounding to minimum versions with yearly scheduled upgrades before the beta every year.

from robotpy-meta.

Currently we bound the versions by year: https://github.com/robotpy/robotpy-meta/blob/main/requirements.txt

from robotpy-meta.

Updated description with what I see as the three options we have now.

from robotpy-meta.

So I think it depends on the release strategy that you guys are using. If everytime wpilib pushes a release, lets say a patch version, then u expect to update robotpy and push a new release then I would pin the exact version of wpilib and other dependencies. If ur goal is to somewhat independent of wpilib releases and only push a release when there are notable changes either to wpilib or robotpy then doing a range like ur doing makes sense.

The issue with the range is it makes it difficult to install a very specific version of robotpy with an expected version of wpilib as I could install an older version of robotpy and get a newer version of wpilib than originally released. This may be good or could be bad since it is possible to have unexpected results possibly.

from robotpy-meta.

Well my question is what release strategy should we use?

I do feel like I'm leaning more towards pinning, specifically because it's hard to install a specific version of robotpy.

from robotpy-meta.

I would agree that I think that pinning is definitely the best thing. It provides the most predictable results.

from robotpy-meta.

I have decided to pin allwpilib based packages, and keep vendors in a year-based bounded range.

from robotpy-meta.