Build failures in Hackage matrix

[bergey]

Recent versions of diagrams-lib have compile-time failures on http://matrix.hackage.haskell.org/package/diagrams-lib, where I would expect "no install-plan". It's easy to tighten up the base constraint. We should be more careful about this in the future, when we remove old GHC versions from travis.

I expected that cabal would decide "no install-plan" because other-extensions includes LambdaCase. This comment seems to say that's the expected behavior. Does anyone know why cabal instead issues a warning and keeps going?

I'd like to leave this open until I have time to retroactively tighten the bound in Hackage revisions.

I'd also like suggestions of how tooling could make this sort of maintainance easier.

[byorgey]

I tightened the lower bound for base, and also went through and retroactively tightened the base lower bound for 1.3.0.4 - 1.3.1.3 on Hackage.

This is definitely going to happen again at some point, though. I am unsure of what sort of tooling we could use to make this easier.