TuringLang / Bijectors.jl

Implementation of normalising flows and constrained random variable transformations
https://turinglang.org/Bijectors.jl/
MIT License
199 stars 32 forks source link

Update Github actions to run tests on Julia 1.6 #229

Closed torfjelde closed 1 year ago

torfjelde commented 1 year ago

Now that Julia 1.6 is the latest LTS, we should be running tests on 1.6 rather than 1.3.

devmotion commented 1 year ago

IMO we should test also the oldest supported version. I've seen too many examples of people accidentally breaking support for older versions when they were not tested.

devmotion commented 1 year ago

BTW I saw the same test error this morning in AbstractDifferentiation: https://github.com/JuliaDiff/AbstractDifferentiation.jl/runs/7891980707?check_suite_focus=true#step:6:400 Maybe some issue with Julia 1.8?

torfjelde commented 1 year ago

IMO we should test also the oldest supported version. I've seen too many examples of people accidentally breaking support for older versions when they were not tested.

Agreed! But should we maybe just drop 1.3? We don't support it in any of the other packages AFAIK.

devmotion commented 1 year ago

My take in similar cases has been so far to not touch the lower bound as long as it does not increase maintenance burden (special handling, debugging etc.) and does not slow down development (preventing new features etc.). I guess here it would be fine to support 1.3 for now. But I am fine either way.

torfjelde commented 1 year ago

Gotcha :+1:

I'm going to go with just dropping testing on 1.3 for two reasons:

  1. It heavily reduces test times.
  2. I'm currently making a lots of changes to Bijectors.jl and will making use of features no longer available on Julia 1.3.
devmotion commented 1 year ago

I'm going to go with just dropping testing on 1.3

OK, in that case it would be good to bump the julia compat to 1.6 right away to avoid tagging a release that unintentionally breaks < 1.6.