My default view isn't the repo view, so the old link sent me to the
timeline view rather than the contributor documentation.

This has caught enough bugs where it's probably justified to always build it,
even if there is no `head.hackage` patch.

Krzysztof Gogolewski authored 3 years ago and Matthew Pickering committed 3 years ago

This reverts commit 9b21848a.

The bug has been fixed.

Ben Gamari authored 3 years ago and Matthew Pickering committed 3 years ago

Fixing #39 with https://github.com/mpickering/ghc-artefact-nix/pull/14.

Not doing so causes issues for packages like `Cabal`, whose patches conflict
with the `installed` constraints.

In theory, this should decrease the likelihood of picking bizarre build plans
involving boot packages that differ from what comes bundled with the version of
GHC being used.

Previously, the `singletons` package included all of what is now part of
`singletons-base`. To ensure that we are still testing this code on CI, make
`singletons-base` the `extra_package`.

This reverts commit fea2790a.

This reverts commit 3bbe1d2b.

Matthew Pickering authored 3 years ago and Ryan Scott committed 3 years ago

ghc@25977ab5 removes the `loadModule` function, which `doctest` was using.
However, there doesn't appear to be a good reason for `doctest` to be using
`loadModule` in the first place, since it only cares about the parsed module.
As a result, we can simply comment out the use of `loadModule`.

Fixes #30.

Co-authored-by: Ryan Scott <ryan.gl.scott@gmail.com>

This is a workaround until a new 9.2 bindist containing
ghc@e17361f2 is available.
See the discussion in
ghc!6066 (comment 371762).

Make sure that `extra_package Agda` doesn't accidentally pick up a
later version of `Agda` than what is currently patched.

Ben Gamari authored 4 years ago and Ryan Scott committed 4 years ago

Co-authored-by: Ryan Scott <ryan.gl.scott@gmail.com>

This ensures that we don't reuse an out-of-date tree if the user re-runs
after modifying a patch.

Unfortunately the interface [1] we were using previously will not
provide an artifact until a pipeline has successfully finished on the
requested branch.

This partially reverts commit 00db7a0d.

[1] https://docs.gitlab.com/ee/api/job_artifacts.html#download-a-single-artifact-file-from-specific-tag-or-branch