The generated file sets its own directory as the source, including the
generated file itself, which causes rebuilds when that file is
reformatted. We can avoid this by overriding the source with a filtered
version and using that throughout the tests.
See https://github.com/NixOS/nixpkgs/pull/320572 for more context
This reverts commit 81c06bc609.
Reason for revert: This change breaks the
`tests.haskell.cabalSdist.assumptionLocalHasDirectReference` test which
relies on checking for the test source store path in the resulting
derivation files. 81c06bc609 did not
account for this in the change (though it should be possible).
The current version of linear (1.22) has incorrect constraints in its
cabal file (which has been corrected in a revision), so it is causing
this test to fail.
This commit just switches to another arbitrary package (cereal) for the
tests.
This commit does the following two things:
1. Changes the Haskell package used in the extraDependencies test from
releaser to conduit. The reason for this is that conduit is more
of a "core" package in Haskell, and it is more likely to always be
working. (If conduit is not compiling, then large chunks of Hackage
won't be working.)
2. Tighten up the GHCi test to actually use the package from
extraDependencies. It appears that GHCi can fail to import the
package from extraDependencies, and that doesn't cause GHCi to
return an error code, which means the test isn't actually testing
anything.
This new change makes it so that if the package from
extraDependencies is actually not included for some reason, then
this shellFor test should fail.
With a recent hackage update, turtle stopped compiling on ghc94. This
commit changes the tests.haskell.incremental test to use the temporary
package instead of turtle.
This allows packagePlatforms to pick up on the overall supported
platforms and schedule builds on Hydra for more than the evaluation
platform (usually x86_64-linux).
Since the rust writer doesn't seem to get fixed on darwin, we'll just
wrap the haskell writer test in our own derivation (which is possible
since tests.writers exposes a bunch of internals via passthru) and
expose it via tests.haskell which are already in mergeable.
Finally a way to test the (hopefully) working haskell writer on darwin
again!
Contrary to database-id-class, linear is part of stackage and actively
maintained, so the test is less likely to fail due to version
constraint issues as it is currently.
