Opinion: Is it time to stop testing with opam < 2.2

Ecosystem
,
3

In my opinion the article mixes up 3 different things:

  • archive-mirrors not having the right priority with opam < 2.2: this could be simply added to opam-repository’s repo file today. I don’t think anyone serious would be using a local mirror with opam 2.1. If someone is, this could be simply warned as a custom announce if the user is using opam < 2.2 but i’m not sure it’s necessary.
  • Stop testing opam-repository with opam < 2.2: this one i don’t think is a good idea just yet. Sadly opam 2.1 is still in use in the wild (see repology) and testing should take into account what users actually use in my opinion. I don’t really have a good answer for the solver timeout with opam 2.0 but maybe errors like that could be ignored since 2.0 is the least used.
  • Stop generating docker images with opam 2.0 as default + the binary for every major opam versions: this is your choice if you want to change the image matrix. My 2 cents is that having the full matrix of images with just the latest version of opam (+ latest beta and master) is reasonable, but there should at least be one <distribution>-opam-<version> generated for the latest versions (LTS & latest) of the compiler for continued testing on opam-repository (see point 2). Given this would be a breaking change, the naming system could maybe be changed or something like that, so that users of those images don’t get their setup broken out of nowhere.
3 Likes