Module re-exports are handled incorrectly

[feuerbach]

Contrary to its documentation, Haddock.Interface.Create.moduleExports incorrectly assumes that for a module re-export all of its symbols are re-exported.

As an illustration, http://hackage.haskell.org/packages/archive/test-framework/0.8/doc/html/Test-Framework.html shows constructors of the Test data type, although it is exported as an abstract type.

[SimonHengel]

It would be interesting whether this was recently introduced or whether it was incorrectly implemented in the first place. Do you have any insights here?

[feuerbach]

No, it just was never fully implemented.

commit fb43b30e5d2e04d0b6e997055b377cf2d67486d9
Author: David Waern <david.waern@gmail.com>
Date:   Sun May 22 21:40:21 2011 +0000

    Break out fullContentsOf, give it a better name and some documentation
    
    The documentation describes how we want this function to eventually behave,
    once we have fixed a few problems with the current implementation.

[SimonHengel]

Thanks for digging this up. I guess it would be useful to add the note in the commit message to the actual code. If you feel like hacking on this, a patch would be very appreciated ;)

[feuerbach]

I might do eventually, but my work queue is already rather large atm :)

Also, (whoever will be working on this), note, that semantics of exports is a bit tricky and easy to get wrong. See e.g. http://hackage.haskell.org/trac/ghc/ticket/7469

So, if possible, use functions from the GHC API to do export and import resolution. I don't know whether such functions exist — could someone knowledgeable about the GHC API comment on this?

[SimonHengel]

No idea, maybe ask on glasgow-haskell-users?