search

brendanhay / amazonka

A comprehensive Amazon Web Services SDK for Haskell.
https://amazonka.brendanhay.nz
Other
605 stars 226 forks source link

amazonka base revisions prevent building in Stackage Nightly! #988

Closed juhp closed 6 months ago

juhp commented 6 months ago

It is seems amazonka packages got revised to base < 4.19 on Hackage: see for example https://hackage.haskell.org/package/amazonka-2.0/revisions/

Given that is was building is this really correct? Maybe the intention was base < 4.20??

cc @JackKelly-Bellroy ?

endgame commented 6 months ago

Yep, that's intentional According to Base Package (HaskellWiki), base-4.19 corresponds to GHC 9.8, and if the linked issue is correct, only ~20 of the ~330 amazonka-* packages were compiling with GHC 9.8. Missing from that list is amazonka itself (necessary for user programs to make AWS calls in an ergonomic way) as well as bindings for many core AWS services like S3, Lambda, EC2, and IAM. Any service which has a record that shares a field name with any other record in the same service will fail to compile because GHC 9.8 changed (for the better) where you need to place {-# LANGUAGE DuplicateRecordFields #-}. See #969 for the original Amazonka bug, #972 for the PR to fix the generator, and #973 for the regeneration of service bindings.

The bounds revisions therefore seem largely correct. If you think that re-revising bounds will improve things for your workloads (should I be raising the base upper bound on just the packages listed in your linked Stackage issue?), then I can raise the upper bound for the 2.0 releases on Hackage.

juhp commented 6 months ago

Okay thanks for responding so fast: apologies looks like I probably "pulled the trigger too fast" - you are right it's not that many packages...

I just wasn't sure what was going on, so the detailed explanation is most helpful. Again sorry for pointing a finger at you 😔

juhp commented 6 months ago

I removed the accusatory sentence from the description, sorry again - I should have checked more carefully on the status: though I couldn't see any obvious recent ticket here immediately.

I know time is limited for all of us: a heads-up or stackage removal could have helped too - anyway thanks again for the quick response and correcting me.

endgame commented 6 months ago

I appreciate that mate, thank you. I've edited out my prickly response, too, so we can call it even.