Add Semigroup Form instance

[RyanGlScott]

This is needed to fix the build on GHC 8.4, where Semigroup is now a superclass of Monoid.

This declares a dependency on semigroups, but http-api-data was already transitively depending on semigroups via attoparsec.

[phadej]

Attoparsec (and IIRC all other deps) have GHC-conditional dependency on semigroups: https://hackage.haskell.org/package/attoparsec-0.13.2.2/dependencies

[RyanGlScott]

OK. Should I interpret that statement as "please change the semigroups dependency in http-api-data.cabal to be conditional as well"?

[phadej]

And we need to bump lower bound on unordered containers, don't we? https://hackage.haskell.org/package/unordered-containers-0.2.8.0/changelog

[RyanGlScott]

...erm, I don't know? I'm not affiliated with unordered-containers in any way.

[RyanGlScott]

After getting some sleep, I realize that I might have wildly misinterpreted #68 (comment). I believe what you're asking is for me to change http-api-data.cabal such that:

• The dependency on semigroups is conditional on if !impl(ghc >= 8.0).
• The lower version bounds for unordered-containers is 0.2.6.

Can you confirm if this is correct?

[phadej]

@RyanGlScott yes! sorry for being cryptic myself, I hadn't enough sleep myself!

[RyanGlScott]

Hm, the build is apparently failing due to this line:

http-api-data/.travis.yml

Line 109 in b61b152

- if [ $HCNUMVER -eq 70804 ]; then cabal new-build -w ${HC} --disable-tests --disable-benchmarks --constraint='attoparsec ==0.13.0.1' --constraint='attoparsec-iso8601 ==1.0.0.0' --constraint='bytestring ==0.10.4.0' --constraint='containers ==0.5.5.1' --constraint='hashable ==1.1.2.4' --constraint='http-types ==0.8.6' --constraint='text ==1.1.1.3' --constraint='time ==1.4.2' --constraint='time-local-compat ==0.1.1.0' --constraint='unordered-containers ==0.2.5.0' --constraint='uri-bytestring ==0.1.7' --constraint='uuid-types ==1.0.2' all; else echo skipping...; fi

which hardcodes a version number for unordered-containers that is older than 2.6.0. Is there a particular motivation for using this version number?

[phadej]

@RyanGlScott yes, I want to check lowerbounds, and seems I have to check them on newer GHCs too, as that check didn't catch incorrect lowerbound (as lowerbounds were tested only on GHC-7.8.4 and GHC-7.10.3)

[RyanGlScott]

OK, I'll try changing those lower bounds to 0.2.6.0.

[RyanGlScott]

...oops, I just noticed that this may be superseded by #69 now.

[phadej]

Sorry for not notifying, merged as part of #69