Doesn't build with GHC 7.10

[robdockins]

Sadly, relaxing package upper bounds doesn't just work:

$ ghc --version
The Glorious Glasgow Haskell Compilation System, version 7.10.1

$ cabal install bitset==1.4.8 --allow-newer
Configuring bitset-1.4.8...
Building bitset-1.4.8...
Preprocessing library bitset-1.4.8...

src/GHC/Integer/GMP/TypeExt.hs:16:8:
    Could not find module ‘GHC.Integer.GMP.Prim’
    Use -v to see a list of the files searched for.
Failed to install bitset-1.4.8

[miguel-negrao]

Is it to be expected that at some point this package will be upgraded to use version 1.0.0.0 of integer-gmp ? It would be useful to know what are the intentions.

[superbobry]

I haven't looked into the sources of the new integer-gmp yet, but since it's complete re-write my guess is that a backward-compatible migration might take some time.

[NCrashed]

I have ported the library to 7.10.2, but dropped backward compatibility (I need to checkout henet package if it fits my needs). Perhaps, someone could reuse the work: https://github.com/Teaspot-Studio/bitset

[superbobry]

Cool! I think backward compatibility is just a question of free time and some ifdef hackery. Feel free to submit a PR.

[treeowl]

@superbobry, my big question is why bitset reimplements instance Bits Integer at all. If there's an efficiency problem with what base does there, shouldn't that be pushed upstream? And if there isn't (any longer) such a problem, shouldn't we strip out all the extra code? This whole thing seems strange. One issue I know about is that some Bits instances in base fail to INLINE some things they should. It would be much nicer to work around this problem using GHC.Exts.inline, which for non-GHC implementations can be defined as id, instead of digging around in Integer internals.

[treeowl]

Ahh..... I see. https://ghc.haskell.org/trac/ghc/ticket/7860 is still open after all these years.