On Tue, Aug 05, 2008 at 12:18:55PM +0100, Duncan Coutts wrote: > On Tue, 2008-08-05 at 00:22 +0100, Ian Lynagh wrote: > > On Mon, Aug 04, 2008 at 10:40:58PM +0100, Duncan Coutts wrote: > > > On Mon, 2008-08-04 at 16:33 +0100, Simon Peyton-Jones wrote: > > > > > > > > John discovered (we think) that a new feature of the build system is > > > > that all GHC's source modules must be listed by the developer in > > > > ghc.cabal. > > > > > > Yes. > > > > Although it's not actually new; before you had to list all the modules > > in package.conf.in. With the old system I suspect it wouldn't go wrong > > until a GHC API client tried to import a module that you'd left out, > > though. > > > > As Duncan says, Cabal will eventually be fixed to fail even earlier. > > Or rather, not fail at all! :-) > > The dep tracking will make other-modules optional as we'll be able to > discover them automatically.
OK, so that'll get us the old behaviour. I'm not convinced that that's better than failing for the ghc case, as (currently, at least) everything is exposed. Does that mean that everything will have to be preprocessed when doing sdist, so that you can discover the deps? What if there are conditional imports? Thanks Ian _______________________________________________ Cvs-ghc mailing list [email protected] http://www.haskell.org/mailman/listinfo/cvs-ghc
