In preparation for ghc/ghc#23719

Specifically:

* ghc/ghc@7902ebf8
* ghc/ghc@4e5ff6a4

The upstream package has cut a new release that is compatible with HEAD: enummapset-0.7.2.0

Bryan R authored 1 year ago and Bryan R committed 1 year ago

I was honestly confused by the names when I first encountered them. What
is being built? What is being tested? Why is it called 'foo-pipeline',
isn't that redundant? (No)

Bryan R authored 1 year ago and Bryan R committed 1 year ago

before_script and script just get merged, so unless we're setting some
default somewhere we don't need it. And in fact, every user of .build
was doing the same thing, so we can just do it in one place: .build.

Fixes ghc/ghc#24183.

I believe the trigger job will fail if there are no jobs in the
downstream pipeline, so it needs the rule anyway. Once it's there, the
rules on the jobs are redundant.

Again, this will let us split up jobs in upcoming commits.

This breaks the dependency between job descriptions and lets us start
putting jobs in different files.

cabal-install might put files either into .cabal or .cache/cabal depending on whether the .cabal dir exists. let's just try both directories for maximum compatiblity

Teo Camarasu authored 1 year ago and Matthew Pickering committed 1 year ago

This is to avoid this bug: https://github.com/NixOS/nix/issues/9142

Deal with re-export of Data.List.List

Suport ghc/ghc!11426

This change was introduced in
ghc/ghc!11367

* Use branches for nixpkgs and all-cabal-hashes, so they can be updated
  with `nix flake lock --update`
* Remove old `default.nix` file.

The fixes have been backported, servant should work in 9.4.5, hgeometry in 9.8.1.