fails to build against HP

[bergey]

Opening this issue to give a place to discuss the current round of build failures. For example:

https://travis-ci.org/diagrams/diagrams-lib/jobs/25631421

One option is not to care; a normal cabal install will just install a new attoparsec, which won't immediately break anything. But we should at least discuss this before releasing 1.2.

As far as I understand:
lens pulls in aeson-0.7.*, which pulls in attoparsec. HP has attoparsec-0.10.4, which limits us to aeson<0.7.0.5. In practice, 0.7.0.4 has enough upper bounds to actually work. The missing upper bounds mean that Cabal always prefers scientific-0.3, and versions of aeson that can't actually build that way.

Constraining aeson to 0.7.0.4 allows travis to build on the Platform: https://travis-ci.org/diagrams/diagrams-core/builds/25829016. We could of course add this in our .travis.yml instead.

[bergey]

I just pushed the aeson constraint to diagrams-travis