# Cabal meeting (25/04/2024)
[Previous meeting notes](https://hackmd.io/c1yOln1fR_K5WiEhhPsUlQ?both).
## Current affairs
- *The neverending Cabal 3.12 release* (Francesco A.): I arrived at point [C.5 Publishing the artifacts](https://github.com/haskell/cabal/wiki/Making-a-release#c5-publishing-the-artifacts). I suspect I need access or someone to upload artifacts for me. (~~Preliminary step: mutually sign key with Mikolaj.~~ Done!)
- [[#9729](https://github.com/haskell/cabal/issues/9729)] 3.12.0.0 release / checklist for GHC 9.10
- **action**: ask Ben whether he prefers tags or hashes, make the instruction more precise, continue release process (f-a).
- [Mikolaj] Shall we do a pre-release, named 3.12.0.0, of cabal-install for (almost) free from the same tag that Cabal is released from? It's to our advantage that users may be fooled it's a real release (usually our pre-release names end with a date) and for once install a pre-release and give us any feedback.
* we've been getting this anyway, from HLS users
* **action:** Mikolaj will attempt it and see how much effort it is and at least update some pre-release docs
- [Artem, [#9914](https://github.com/haskell/cabal/pull/9914)] CI update for GHC 9.10
## Review needed
<!--
We need to take a look at these PRs and help them move on.
Urgent, of various levels of importance
-->
## Food for thought
<!--
We want to start the thinking process amongst ourselves
for things that will take time to unfold.
May be important, but not urgent
-->
[geekosaur, [#9934](https://github.com/haskell/cabal/issues/9934)] Splitting up CI to speed up dev
* Just floating this now; I don't expect anything to happen until after cabal-install 3.12.1.0 release.
[geekosaur] I was asked last meeting to look into C++ modules.
* `clang` appears to support them fully, including separate compilation of CMIs (these are roughly equivalent to Haskell's `.hi` files).
* `g++` has only partial support, and does not support separate compilation of CMIs. It also claims that separate compilation is transitional, unlike `clang` which appears to consider it a standard feature.
* The above is significant because separate CMIs would require specific support in both `ghc` and `cabal`, but would remove the need to compile modules before normal C++ code in `c++-sources`.
[romes] I stress tested private dependencies and that uncovered a really bad performance issue in the closure property check. I've addressed it, and the stress test shows negligible overhead for checking the closure property (which now takes +-70ms).