gistfile1.txt

Downloading semigroups-0.16.2.2...
Configuring semigroups-0.16.2.2...
Building semigroups-0.16.2.2...
Preprocessing library semigroups-0.16.2.2...

src/Data/Semigroup.hs:105:8:
    Ambiguous module name ‘Numeric.Natural’:
      it was found in multiple packages:
      nats-0.1.2@nats_1SlQ91QfEGVBnpiEUJoePq base-4.8.0.0
Failed to install semigroups-0.16.2.2
cabal: Error: some packages failed to install:
MemoTrie-0.6.2 depends on semigroups-0.16.2.2 which failed to install.
implicit-0.0.3 depends on semigroups-0.16.2.2 which failed to install.
semigroups-0.16.2.2 failed during the building phase. The exception was:
ExitFailure 1
vector-space-0.10.2 depends on semigroups-0.16.2.2 which failed to install.
void-0.7 depends on semigroups-0.16.2.2 which failed to install.

Numeric.Natural was in nats, but is now (in the new ghc you are using) in base. This is a little unusual. Thus recent nats-1.etc is a empty package for ghc-7.9+ -- the module is installed only where base (and thus ghc) are older. See the last paragraph in http://hackage.haskell.org/package/nats-1/nats.cabal The difficult question is why, with your shiny new ghc, something directed you to install nats-0.1.2. If something did, it probably has out of date bounds. If you do ghc-pkg unregister nat you will be warned what it might break, and thus what might have asked for nats-0.1 to be installed. You certainly cannot have a version of nats < 1; they will all overlap with the newest base.

Oh wait I see you've discussed this with cartazio et al elsewhere.

The main trouble here is in any case your super-new ghc. Hackage isn't entirely in sync with it yet.