[darcs-devel] Re: darcs patch: Add Cabal files (not feature complete).

[David Roundy]

On Mon, Apr 09, 2007 at 10:00:20PM +0200, Benedikt Schmidt wrote:
> "Zachary P. Landau" <kapheine at divineinvasion.net> writes:
> 
> >> Why Cabalisation?  I'm not too sure.  It does make it easy to
> >>  - express package dependencies (e.g., mtl, parsec, QuickCheck)
> >>  - expose a few modules and create a libdarcs
> >>      In fact, we might consider having two cabal files: one for libdarcs
> >>      and the other for darcs proper; might make the install process a
> >>      bit more complicated, however)
> >
> > It's always nice to have programs using a standard packaging mechanism.
> > I think this is a good idea.  It might help package managers for Linux
> > distributions, as well (although by now most of the major ones have
> > already made a package for darcs).
> 
> Another advantage is that there are many haskell tools that rely on the
> cabal file for build-info (cabal-test, visual-haskell, shim).
> Btw, someone asked about a faster compile-edit cycle a few months ago,
> now that we have a cabal-file, the emacs shim-mode [1] typechecks mosts
> darcs files in a few seconds. So I would really like this patch to go
> into unstable.

I should perhaps pipe up that I've no objection to cabal use, although I
haven't actually looked at the patch.  Does the cabal support replace the
makefile? That'd be great, it'd be a bummer to have a distinct mechanism
for building documentation, and I don't know how flexible cabal is, in
terms of serving as a replacement for make.  Or is the cabal stuff actually
rather light-weight, in which case it wouldn't be burdensome? Also, does
cabal allow us to represent the fact that darcs depends on unix on unixy
systems, but on win32 in windowy systems, and things like that?
-- 
David Roundy
Department of Physics
Oregon State University