On 2015-01-27 at 18:25:21 +0100, Gershom B wrote: [...]
We could perhaps also insert a warning encouraging people to import Data.List qualified, either in this GHC of the next release. This would allow BBP to proceed in a relatively breakage-free way in the future without needing to generalize the Data.List module.
Ironically, the AMP-related warnings in GHC 7.8 weren't very effective at getting noticed timely in order to have Hackage ready for the actual AMP-changes occurring in GHC 7.10. By far the most breakages[1] I encountered (and reported) over the last weeks in popularly-depended-upon packages were due to missing Functor/Applicative instances. It'd be interesting to know why those warnings weren't acted upon during 2014. Otoh, I don't have numbers how many packages were actually fixed up *thanks* to the GHC 7.8 warnings... [1]: FWIW, the AMP is a serious breaking change due to its nature of introducing a superclass constraint *incompatible* with H2010 as H2010 doesn't even know anything about the Applicative class to begin with... The BBP, on the other hand, in its incarnation as of GHC 7.10.1RC2 was designed to be a smooth upgrade path for existing H2010 code as it just generalises type-signatures, and does not even require any CPP to write code that compiles with both, a strict Haskell2010 compiler+base-library, as well with GHC 7.10.1RC2's F/T-generalised base-library. Cheers, hvr