Support GHC 8.10

[sjakobi]

No description provided.

[sjakobi]

CI with 8.10 failed due to haskell-checkers/checkers#44.

[sjakobi]

I have lowered the upper bound on checkers to work around the build failure.

Travis should succeed too now: https://travis-ci.org/github/hedgehogqa/haskell-hedgehog/builds/669930888

[moodmosaic]

@sjakobi, cool 🍻 I just re-run the AppVeyor build.

[sjakobi]

Not sure what's up with AppVeyor…

Could you try deleting its cache?

[moodmosaic]

@sjakobi

Could you try deleting its cache?

There is an option "Save build cache in Pull Requests" which is turned off.

[sjakobi]

Could you try deleting its cache?

There is an option "Save build cache in Pull Requests" which is turned off.

Hm, not sure what to do except try rerunning then.

[sjakobi]

The AppVeyor failures look similar to haskell/cabal#4005. How about turning off builds with GHC 8.0?

[sjakobi]

OK. AppVeyor and Travis are finally both green now.

[jacobstanley]

@sjakobi did you use bin/regenerate-travis.sh for this? I think it's missing the -Wall -Werror that we used to have

[sjakobi]

@sjakobi did you use bin/regenerate-travis.sh for this? I think it's missing the -Wall -Werror that we used to have

Oh no, I wasn't aware of that! It's strange that --copy-fields=all doesn't show up in the incantation in the .travis.yml.

Why don't you use the more conventional cabal.hasell-ci to configure haskell-ci?

[sjakobi]

The Travis job with GHC 8.0 is failing now because of the combination of -Werror=missing-methods with -Werror. Previously it just produced a warning.

I suspect it's a haskell-ci bug to pass the unsupported -Werror=missing-methods flag to a GHC < 8.2.

[sjakobi]

I reported the haskell-ci issue in haskell-CI/haskell-ci#377.

[sjakobi]

I acted on @phadej's excellent advice in haskell-CI/haskell-ci#377 (comment) and improved the haskell-ci incantation in a way that made the regenerate-travis.sh script unnecessary. In the future, a simple haskell-ci regenerate will be sufficient.

Unless haskell-checkers/checkers#44 gets merged very soon, I plan no further changes. ;)

[moodmosaic]

LGTM

[moodmosaic]

Is haskell-ci --copy-field=all cabal.project an alternative? If yes, I think it makes sense to remove this (as you did).

[sjakobi]

Indeed. In general, once the correct incantation is recorded in the comment at the top of the .travis.yml, haskell-ci regenerate should be sufficient.

[moodmosaic]

Good to know 👍 Yes, as per @phadej in haskell-CI/haskell-ci#377 (comment).

[jacobstanley]

Why don't you use the more conventional cabal.hasell-ci to configure haskell-ci?

I wasn't aware, thanks for the changes! 🎉

[jacobstanley]

Note @erikd has updated the template-haskell bounds to < 2.17 on the existing Hackage version hedgehog-1.0.2, so that should now be good to go also.

Thanks all 🎉

[sjakobi]

Thanks for the merge!

This was a bit of a yak shave! ;)