Allow using io-classes 1.8 #101
Conversation
b07784b to
9905331
Compare
9905331 to
b5b4cad
Compare
e0ed8aa to
de4a084
Compare
- update minor versions - update macos
de4a084 to
60c06ce
Compare
There was a problem hiding this comment.
drive-by comment
I hope you're all doing well!
The lower bound on base should arguably be updated if support for ghc-8.10 and ghc-9.2 is dropped
Related to that -- I'm okay with dropping ghc-8.10 support, but ghc-9.2 support seems relatively painless to keep. Dropping ghc-9.2 means that lsm-tree also has to drop ghc-9.2 (which we're happy to keep supporting). Given that fs-api and fs-sim are packages on Hackage, I think we should be careful with dropping support for ghc versions anyway
I tried building latest main (a9900ad) with ghc-9.2 and I realise there is a build error in io-classes:stm, which I suppose is the reason why you dropped support
Configuring library 'strict-stm' for io-classes-1.8.0.0...
Preprocessing library 'strict-stm' for io-classes-1.8.0.0...
Building library 'strict-stm' for io-classes-1.8.0.0...
[1 of 7] Compiling Control.Concurrent.Class.MonadSTM.Strict.TArray ( strict-stm/Control/Concurrent/Class/MonadSTM/Strict/TArray.hs, dist/build/strict-stm/Control/Concurrent/Class/MonadSTM/Strict/TArray.o, dist/build/strict-stm/Control/Concurrent/Class/MonadSTM/Strict/TArray.dyn_o )
[2 of 7] Compiling Control.Concurrent.Class.MonadSTM.Strict.TBQueue ( strict-stm/Control/Concurrent/Class/MonadSTM/Strict/TBQueue.hs, dist/build/strict-stm/Control/Concurrent/Class/MonadSTM/Strict/TBQueue.o, dist/build/strict-stm/Control/Concurrent/Class/MonadSTM/Strict/TBQueue.dyn_o )
strict-stm/Control/Concurrent/Class/MonadSTM/Strict/TBQueue.hs:44:22: error:
• Illegal equational constraint LazyTBQueue m ~ LazyTBQueue n
(Use GADTs or TypeFamilies to permit this)
• In the type signature:
castStrictTBQueue :: LazyTBQueue m ~ LazyTBQueue n =>
StrictTBQueue m a -> StrictTBQueue n a
|
44 | castStrictTBQueue :: LazyTBQueue m ~ LazyTBQueue n
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^...
Error: [Cabal-7125]
Failed to build lib:strict-stm from io-classes-1.8.0.0 (which is required by test:fs-sim-test from fs-sim-0.3.1.0). See the build log above for details.
But this build error is arguably an io-classes bug that should be resolved in a patch to io-classes, either by fixing the build failure or having stricter base bounds to rule out ghc-9.2
|
Thanks Joris!! I noted it down on an issue in the io-sim repository. Let's see what the maintainers decide. We will not publish fs-sim and fs-api until this is resolved one way or another. |
- tighten the lower bound on `base` - tighten the constraint on `io-classes` Follow-up to input-output-hk#101
- tighten the lower bound on `base` - tighten the constraint on `io-classes` Follow-up to input-output-hk#101
- tighten the lower bound on `base` - tighten the constraint on `io-classes` Follow-up to input-output-hk#101
This PR allows using
io-classes-1.8. We have to drop support for GHCs older than 9.4 asio-classes-1.8does not support them.