release progress GHC 9.12.5

Magnus Viernickel <[email protected]> Tue, 26 May 2026 18:14:44 +0200
Newsgroups gmane.comp.lang.haskell.glasgow.user
Message-ID <[email protected]>
Dear GHC release stakeholders,


We are approaching a release candidate for GHC 9.12.5. Currently there's 
only one blocking patch 
(https://gitlab.haskell.org/ghc/ghc/-/merge_requests/15640) that still 
needs to be backported and the test-suite seems to be doing well.

Here are some important notes for this release of GHC 9.12.

1. Our Darwin CI is now running on MacOS version 26 for the aarch64 
runners and on MacOS version 15 for the x86_64 runner. The SDK version 
on the MacOS 15 runner is 26.2, the SDK version on the aarch64 runners 
is 26.4. Note that this may affect bindists.
2. See https://gitlab.haskell.org/ghc/ghc/-/issues/27286 for an 
exhaustive list of MRs that have been backported.
3. If you are a distributor of haskell packages like stackage or 
nixpkgs, I would appreciate testing an RC with you, as soon as it's 
there. Please reach out if that's the case.
4. In ncoordinatio with the cabal developers, the GHC team has decided 
to stop maintaining the V1 ghc semaphore protocol that had some 
fundamental architectural flaws. While this is technically a 
(performance) regression (cabal versions that only support v1 will fall 
back to -j1 even when passed -jsem with 9.12.5) and adds a significant 
feature, it is the most pragmatic approach to fixing the glaring issues 
with the previous semaphore implementation without having to maintain a 
fundamentally broken design. Please check out 
https://gitlab.haskell.org/ghc/ghc/-/merge_requests/15729 for further 
information. We expect that `-jsem` will work out of the box with the 
next released cabal version and 9.12.5.

I expect a release candidate to appear this week, depending on progress 
of the blocking patch.

Please don't hesitate to reach out if you have any questions or remarks.

Best

Magnus

_______________________________________________
Glasgow-haskell-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]