search

jberthold / packman

Evaluation-orthogonal serialisation of Haskell data, as a library
Other
59 stars 9 forks source link

switched failing test suites from detailed-0.9 to exitcode-stdio-1.0 #7

Closed michaxm closed 8 years ago

michaxm commented 9 years ago

... which is to my knowledge still defunct (please correct me, if I am wrong)

michaxm commented 9 years ago

See https://github.com/jberthold/packman/issues/4 or rather the gist linked there: https://gist.github.com/rrnewton/5f8b1db0a3f264ab48b7

jberthold commented 8 years ago

Thanks for your PR. Unfortunately, the situation is a bit more complicated: The three detailed-0.9 tests (alltests, testmthread, quickchecktest) work fine with ghc-7.10, which brings in Cabal-1.22.*. However, they do not compile with ghc-7.8, which comes with Cabal-1.18. If I put a constraint Cabal >= 1.20 (where the LibV09 module exists), the tests cannot be built with ghc-7.8 either, because the library depends on ghc-7.8, which fixes the Cabal version to 1.18.

I would be more inclined to put your PR on a ghc-7.8 branch than to use it for the main line. Would that be enough to work with the library?

I have created a branch wip/fixtests where I will try to split the versions that lead to this conflict. Hope to have more time for this, and other things related to packman early next week.

jberthold commented 8 years ago

@michaxm Could you please test the code from #8 with your setup? (switches to Cabal>1.20, works for me with two different ghc-7.8 installations (one clean, one rather custom) I can adapt your PR to use exitcode with ghc-7.8 and detailed otherwise, but prefer the other variant if it works for you.

michaxm commented 8 years ago

Good news, using cabal-install (and sandboxes) this works for me.

Executing the tests with stack won't work out of the box, because the standard resolution plans are not to be able to find a solution on these constraints - but if cabal can do it, stack should be configurable, too, I might try that on next occasion. And the error message is now a lot more meaningful.

Please use your changes and discard mine, finally a working example with detailed-0.9 :) .