Luckily, all build failures where just related to type checking changes
in GHC 9.0 (and maybe base changes?), so we can just use GHC 8.10.7 with
the Stackage version of packages and keep Yi working. I feel like we
dodged a bullet here, as upstream maintenance doesn't seem to be too
proactive these days.
We need to manually simulate the following conditional in the cabal
file:
if impl(ghc >=9.0)
build-depends: ghc-prim
else
build-depends: hashable >=1.3.5.0 && <1.5
This means we only have to update the llvmPackages attribute in one
place now and should prevent situations like with 8.6.5 where different
versions would be used in the package set compared to the compiler
build.
Drop comments in the configuration-ghc-X.Y.x.nix files as well, since
LLVM version isn't tied to the compiler minor version at
all (e. g. 8.10.2 and 8.10.7 have different support ranges).
* no released version of hackage2nix does support distribution-nixpkgs yet.
* hackage-db 2.1.2 fixes an annoying bug introduced in 2.1.1 and also supports
Cabal 3.4: https://github.com/NixOS/cabal2nix/issues/501
Stackage Nighly recently upgraded their version of hackage-db from 2.1.0
to 2.1.1. 2.1.1 had a compatibility fix for Cabal 3.4 [1]. However it
did not increase the version bound on Cabal nor fails to compile with
Cabal 3.2, so Stackage was able to update it.
Unfortunately hackage-db with Cabal 3.2 causes observable issues [2]
in cabal2nix, so we need to downgrade it for all compilers that still
ship a Cabal version < 3.4.
Also ideally we should update the constraints for hackage-db 2.1.0 and
hackage-db 2.1.1 on hackage. See also [3].
[1]: https://github.com/peti/hackage-db/pull/12
[2]: https://github.com/NixOS/cabal2nix/issues/501
[3]: https://github.com/peti/hackage-db/pull/14
They are not an exposed part of haskellPackages per se, so we shouldn't
list them in nix-env. Additionally this should prevent the failed lldb
build from cluttering our jobset output.
Commit 1998b95adc “haskellPackages:
update default compiler from ghc 8.10.2 to version 8.10.3” broke Agda
because Agda.cabal incorrectly declares a transformers dependency only
for ghc < 8.10.3:
if impl(ghc >= 8.6.4) && impl(ghc < 8.10.3)
build-depends: transformers == 0.5.6.2
if impl(ghc >= 8.4) && impl(ghc < 8.6.4)
build-depends: transformers == 0.5.5.0
if impl(ghc < 8.4)
build-depends: transformers == 0.5.2.0
leading to this error:
src/full/Agda/Utils/Maybe.hs:13:1: error:
Could not load module ‘Control.Monad.Trans.Maybe’
It is a member of the hidden package ‘transformers-0.5.6.2’.
Perhaps you need to add ‘transformers’ to the build-depends in your .cabal file.
Use -v (or `:set -v` in ghci) to see a list of the files searched for.
|
13 | import Control.Monad.Trans.Maybe
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Signed-off-by: Anders Kaseorg <andersk@mit.edu>
Tests did not compile.
Message:
test/Servant/SwaggerSpec.hs:44:14: error:
• No instance for (HasSwagger
(Fragment Int :> Servant.Test.ComprehensiveAPI.GET))
arising from a use of ‘toSwagger’
• In the expression: toSwagger comprehensiveAPI
In an equation for ‘_x’: _x = toSwagger comprehensiveAPI
In the second argument of ‘($)’, namely
‘do let _x = toSwagger comprehensiveAPI
True `shouldBe` True’
|
44 | let _x = toSwagger comprehensiveAPI
| ^^^^^^^^^^^^^^^^^^^^^^^^^^
In preparation of the upcoming 0.6.0 release I wanted to fix hls.
It introduces two new plugin packages, which are not on hackage yet.
I remove apply-refact overrides, because current apply-refact versions
are compatible with all ghcs we support, according to their changelog.
I override more of the hls dependencies globally on the whole package
set, to avoid a lot of duplicate compilations. And because @peti changed
my mind about this being a good practice.
hls now uses a released version of ghcide
