Could we simplify our use of projects?

[philderbeast]

We have 14 packages and 10 (or 11) projects. That's a lot of potential ways to group packages within projects (if sorted - even more if jumbled). Could we use imported groups of packages to simplify this?

$ tree -P '*.cabal' --prune -L 2
.
├── bootstrap
│   └── cabal-bootstrap-gen.cabal
├── buildinfo-reference-generator
│   └── buildinfo-reference-generator.cabal
├── Cabal
│   └── Cabal.cabal
├── cabal-benchmarks
│   └── cabal-benchmarks.cabal
├── Cabal-described
│   └── Cabal-described.cabal
├── cabal-dev-scripts
│   └── cabal-dev-scripts.cabal
├── cabal-install
│   └── cabal-install.cabal
├── cabal-install-solver
│   └── cabal-install-solver.cabal
├── Cabal-QuickCheck
│   └── Cabal-QuickCheck.cabal
├── Cabal-syntax
│   └── Cabal-syntax.cabal
├── Cabal-tests
│   └── Cabal-tests.cabal
├── cabal-testsuite
│   └── cabal-testsuite.cabal
├── Cabal-tree-diff
│   └── Cabal-tree-diff.cabal
└── solver-benchmarks
    └── solver-benchmarks.cabal

15 directories, 14 files

$ tree -P 'cabal.project*' --prune -L 2
.
├── bootstrap
│   └── cabal.project
├── cabal.project
├── cabal.project.buildinfo
├── cabal.project.doctest
├── cabal.project.latest-ghc
├── cabal.project.libonly
├── cabal.project.meta
├── cabal.project.release
├── cabal.project.validate
├── cabal.project.validate.libonly
└── cabal.project.weeder

2 directories, 11 files

[philderbeast]

Anyone know why we need both cabal.project.validate and cabal.project.validate.libonly? Why does the former import: cabal.project.latest-ghc but the later doesn't? This was added in #9330, @ulysses4ever. Do you think we could get by with one cabal.project.validate and delete cabal.project.validate.libonly?

[ulysses4ever]

Yeah, I didn't study it in depth, I just did the minimal necessary chance to make CI support latest GHCs, if necessary, but even that is not guaranteed to work. I really hope we could cut down on our project files...