David Allsopp

Does this need to be done on every platform? It doubles the build work - wouldn’t it be better as a next pipeline stage after the other builds have succeeded...

@tmcgilchrist - had you experimented with this a while back? The major concern - at least for the “main” cluster, is that the trust model for the workers changes as...

Ouch, yes - if there isn't a workaround for here then we should just reject, I agree. I'll open an issue on opam-repo-ci to workaround it.

Oh, there is a possible workaround, remembering horrid things we used to do in OCaml's testsuite - escaping newlines as ` :; \` certainly works here.

The ocaml/opam images on [Docker Hub](https://hub.docker.com/r/ocaml/opam) are primarily intended to support CI systems (in particular, opam-repo-ci for [ocaml/opam-repository](https://github.com/ocaml/opam-repository) and the [ocaml-ci GitHub App](https://github.com/apps/ocaml-ci)). opam, when using Git remotes, expects git...

@avsm - I think these two functions were for obi and aren't being used/tested anywhere. Up to you whether it's worth putting this in, therefore (I only landed on updating...

This didn't need to be merged for the base images - the repo is added there

> This is not persisted in 'native_c_compiler' output of 'ocamlc -config' because that still refers to 'gcc' (although using older GCC to compile stubs "works", but obviously won't help if...

The problem doing this in the base is that it breaks 2.0 - opam 2.1 recomputes the `jobs` variable if it's missing, but opam 2.0 just assumes "1". I think...

I think that the target here would be to have a precise switch invariant - that'll do exactly what the pin did before without triggering #61 (at least I don't...