should haddock.ghc be a sub-repo of ghc?

[Claus Reinke]

>>    http://hackage.haskell.org/trac/ghc/ticket/1886
> 
> I quite like this idea, but it's a substantial upheaval, and therefore more 
> of a long-term proposition.
> 
> In the meantime we could consider the less invasive change: put FastString 
> in place of HsDoc in HsSyn.  No need to pass functions into GHC to do the 
> renaming: Haddock can do it itself since it only needs the top-level 
> GlobalRdrEnv which GHC provides.  I don't think Haddock needs HsSyn with 
> parsed/renamed documentation embeded, since it can keep the documentation 
> in a Name->Doc mapping after it has extracted and processed it.  Sound 
> reasonable?

anything that makes comments easier to get to sounds like 
and improvement. and if the comments end up right in the
AST, i'm not sure what we're missing compared to the ticket?

also, could someone please check why printForUser outputs
syntactically invalid instance code in the ticket example (the
comments in the original source list what goes wrong, but the 
comments were not preserved in the output, of course)?

thanks,
claus