Added stack compatiblility, compile and runs tests with LTS-5.4

[lucamolteni]

A simple stack init generates this stack file that compiles without problem with the latest LTS 5.4

[lucamolteni]

Uhm... The travis build failed, but I can't see really why. Cabal should ignore the presence of a stack file.

[vincenthz]

The tests failed for another reason, which I'm having a hard time to track down;

I don't think there's any benefits to adding stack files to libraries; it's only useful for program where you want everyone to use the same basis of libraries, but for library which are supposed to build in many different package sets, it's probably harmful.

[lucamolteni]

How is it harmful? I'm really interested.

[mgsloan]

As long as your CI is running cabal install, then I don't see what the difference is. Here's my view on why having it in the repo is helpful: https://www.reddit.com/r/haskell/comments/415vne/to_ship_stackyaml_or_not_to/

Just because there is a stack file does not mean you should leave off your version constraints.

We have plans for making stack try different versions - commercialhaskell/stack#1568 . Once that's done, stack will be even better than cabal at version ranges, because it will actually test them.

That said, I don't care much. I'm just commenting because it was pointed out in #haskell-stack

[vincenthz]

To clarify the 'harmful'; I think it send the message that the specific version of stackage that the stack.yaml file use would be preferred compared to what it should really be: many different set of library environment are supported.

Also, I don't think it add any value:

• stack init would generate the same file. there's no extra customisation in it.
• travis would not run any tests against the stack, leaving it for bit-rot
• I rarely/never use stack for general library that are supposed to work in a wide choice of env, leaving it for bit-rot more.

That said, I don't have that much strong filling against it