Re: RFC: include a cabal-install executable in future GHC releases

Am 22.01.2014 13:22, schrieb Roman Cheplyaka:

* Herbert Valerio Riedel [2014-01-22 12:55:53+0100]

...

When I look at http://hackage.haskell.org/package/haddock, there are multiple versions, 2.12.0 and the 4 versions of 2.13.*, which are all declared to work with ghc == 7.6.*, that is, 5 haddock versions compatible with 3 released versions of GHC 7.6. So while it may be bound to a major version of the GHC API, haddock can obviously have more releases than GHC has releases (otherwise it could just carry GHC's version), and can therefore be updated.

Henning could be more specific in his claim that "haddock is bound to the specific GHC version".

What I mean is: The Haddock that comes with GHC has a name like haddock-ghc-7.6.3 and when I (cabal-)install a package with a certain GHC versions then the haddock-ghc according to the ghc version is chosen. If I install a package for different GHC versions then I get a warning that the global haddock database version does not match the one of the GHC version used for the particular package. Now, I become uncertain whether this is an argument pro or contra including Haddock in the binary GHC tarball.