TuringLang / Turing.jl

Bayesian inference with probabilistic programming.
https://turinglang.org
MIT License
2.03k stars 218 forks source link

Drop support for ADTypes 0.2 #2243

Closed yebai closed 4 months ago

yebai commented 4 months ago

ADTypes 0.2 doesn't support AutoTapir yet.

yebai commented 4 months ago

The standard approach for fixing a bug is through a new release instead of reversing a release, right?

coveralls commented 4 months ago

Pull Request Test Coverage Report for Build 9302197105

Details


Totals Coverage Status
Change from base Build 9301631361: 0.0%
Covered Lines: 0
Relevant Lines: 1529

💛 - Coveralls
devmotion commented 4 months ago

The standard approach for fixing a bug is through a new release instead of reversing a release, right?

People seem to have quite diverse (and strong) opinions on this matter 😄 But some maintainers of the registry are definitely not happy about yanking releases. However, as stated in the README

There is however, a special category of bugged releases that can not be resolved by having a patch release. These also may to be resolved by yanking. That special category is when the compat bounds have been set too wide. i.e. say v2.10.0 was released using a feature not on julia v1.6 but the compat entry for julia was not raised in the release. In this case releasing a v2.10.1 with the corrected julia compat would not solve the issue as on julia v1.6 Pkg would still resolve the broken v2.10.0, and as a minor bump, reverting the code changes would not be valid in a patch bump. In this case one may either submit a PR to retroactively adjust the compat bounds of previous versions (best user-facing results, but slow and error-prone to implement) or yank the offending release.

yebai commented 4 months ago

I wasn't aware of the yanked option, but it seems appropriate for this purpose. I created a PR:

https://github.com/JuliaRegistries/General/pull/108024

torfjelde commented 4 months ago

Great! Approving now then.