Better fix for #19 about fgl HOT 3 CLOSED

You added a minimum bound to transformers in your second code block; was that meant to be for containers instead?

I'm not sure I like the flag approach (though I can see now why the version of containers should depend upon whether deepseq is being used or not; I just blindly went by what GHC has rather than being compiler-agnostic).

I suppose the upper bound (if any) on containers would then go in the main build-depends block (or should it be taken out of there and only mentioned within the flag bit)?

from fgl.

You added a minimum bound to transformers in your second code block; was that meant to be for containers instead?

Yes, sorry... I've fixed the example

I'm not sure I like the flag approach (though I can see now why the version of containers should depend upon whether deepseq is being used or not; I just blindly went by what GHC has rather than being compiler-agnostic).

Just because GHC ships with a specific containers version doesn't guarantee cabal will pick that version for you (the solver has a soft preference for installed packages though). In fact, here's a compatibilty matrix:

I suppose the upper bound (if any) on containers would then go in the main build-depends block (or should it be taken out of there and only mentioned within the flag bit)?

I'd put a global upper bound in the main build-depends as that's less error-prone
(and btw, the build-depends property is a conjunctive Monoid)

from fgl.

Whilst this works, it results in older GHC's having newer containers built on them on Travis-CI; making the default False results in containers being downgraded on 7.4.2.

Any idea how this can be resolved?

from fgl.