# Cabal meeting (20/06/2024)

[Previous meeting notes](https://hackmd.io/ECxNIHaUTfaZFYHDVX3zSg?both).

## Previous meeting wrap-up

## Current affairs

- 3.12.1.0 release is going strong. Current step changelogs.
- Mikolaj says: we noticed that https://github.com/haskell/cabal/pull/10128 (Cabal: Add flag to ignore build tool dependencies) may be not merged in time for the 3.12.1.0 release, since backports have already been closed, but if so, Hadrian can temporarily work around. Nobody disagreed that Cabal is a fine place to solve this in the long term (and then GHC will bump it's Cabal tag and include it and remove any workarounds). So we agreed to proceed with the PR, discussing there and resolving eny remaining sticking points, in particular the points from Andrea, and then merging.
- Gershom just protested against the proposed solution on Matrix, fwiw
- Artem asks Brandon and Mikolaj to create a couple bullet-list points summarizing their struggle with changelog.d and meditate what can be improved for the future
- documentation should warn not to remove `changelog.d/config` when removing old `changelog.d` fragments
- future work: add a list of valid packages to `changelog.d/config` and validate changelog fragments and command line against it (and near-future work: ignore case in package names)
- we had a backport which accidentally restored some deleted changelog fragments ([#10005](https://github.com/haskell/cabal/pull/10005)). We should be more careful about this.
- Mikolaj says that yes, but actually that was for the best (or maybe even intentional), because the PR fixed some things that fixed some little changelog files so we had the opportunity to review and fix them in the existing big changelog file (in particular, PR numbers were missing in some little files --- have we fixed that in the big changelog?)
- yes (Brandon)

## Review needed
* [#10124](https://github.com/haskell/cabal/pull/10124) cabal 3.12.1.0 release notes
- **action**: done (f-a)

* Javier: The `default-package-bounds` [PR #9570](https://github.com/haskell/cabal/pull/9570) has been updated, it now works also in the presence of `--allow-newer`. What should be the next steps to merge this change?
* **action** Matthew will have a look (thanks Matthew!)

* Javier: CI stuff, not urgent so only if time permits:
* I would like to mention why I'm making [#10125](https://github.com/haskell/cabal/pull/10125), just for people to be aware.
* Consensus: it is ok!
* Would it be OK to disable Windows-GHC9.0.2 due to [#10114](https://github.com/haskell/cabal/pull/10114)?
* Consensus: just pass the flag if not on CI, instead of changing CI.

## Food for thought

* Hécate: Should we inform the GHC team that more things need to be thought-out / made explicit in their planning when new non-reinstallable packages are created / removed?
* For context, [#10093](https://github.com/haskell/cabal/pull/10093) and [#10087](https://github.com/haskell/cabal/issues/10087) happened 2 weeks ago at a point where we were already asked (with reason) why we had not released cabal-install-3.12.0.0. Had we released earlier, this would have been a bad release and an emergency patch would have had to be deployed.
* (Brandon) I keep wanting to propose to GHC that non-reinstallable libraries and dependencies (e.g. `ghc` makes most of the bundled packages non-reinstallable if used) be listed in a file in the settings directory for use by cabal and other build tools. The main problem is that this needs to be coordinated between ghc, cabal, possibly stack, and (for older versions which would need the file to be retrofitted) ghcup, and details need to be ironed out.

* Hécate: I just saw Firefox [127.01's Release Notes](https://www.mozilla.org/en-US/firefox/127.0.1/releasenotes/) and they have an "Unresolved" section, which I imagine are the unfixed regression. It would be nice of us to advertise them more transparently, so that people who rely on fixes are informed that this release (minor or patch) doesn't include the fix. The feedback will help us prioritise (by fixing ourselves are calling for volunteers to help out).
- **action**: (f-a) modify the wiki, check wording with Hécate and Mikołaj.
- will need to exercise judgement: "regression" labels are a good first approximation for what could go in this section but there should be some subjective prioritization, perhaps.

* Matt Pickering: we need a solution to the ghc vs. alex/happy issue ([#10128](https://github.com/haskell/cabal/pull/10128))
- We are proceeding with the PR from Rodrigo
- Mikolaj says: also, look above at the summary of our discussion about this earlier during the meeting.

* Kristen: Private dependencies ([#9743](https://github.com/haskell/cabal/pull/9743)).
- release it as experimental?
- Well-Typed can help with reviewing
- There will be a blog post, hopefully.
- Performance is a sticky issue. At least, if private deps are not allowed on Hackage, there shouldn't be much concern about performance.