[Gtk2hs-devel] cabalized Gtk2Hs

Duncan Coutts duncan.coutts at googlemail.com
Fri Mar 26 16:28:45 EDT 2010

On Sat, 2010-03-13 at 18:31 +0100, Axel Simon wrote:
> Dear all,
> this is a first announcement of cabal packages for Gtk2Hs. This is  
> meant for all those who can spare five minutes to check the packages  
> and to improve them.

This is great Axel. I'm glad you've taken this on (and a bit sorry I
never got round to it).

Can you let us know what annoyances you've hit in Cabal, so we can make
sure they are all filed properly in the Cabal trac and so we can
prioritise them appropriately.

I expect your main problems are with .chs and .chi files. Both the fact
that there is no proper dependency tracking, and the fact that .chi
files from one package are not installed, yet needed when processing
other packages (how do you get around that btw?).

Another thing that I bet is annoying is that cabal-install isn't great
for building local collections of interdependent packages. One wants a
feature to be able to tell cabal that there are several locally unpacked
packages (ie the gtk2hs source tree, one package per subdir) and have
cabal build them in order.

Similarly, once you've generated tarballs for each package, cabal isn't
yet able to easily be told to install a bunch of tarballs and do

I have plans to make cabal-install accept more notions of package, not
just the local dir and package names from hackage but also specific
dir(s) or specific tarballs.

On the "gtk2hs-buildtools" front, that seems like a reasonably way to
go. I hope we can pull the c2hs fork out of that at some point however
and use the mainline c2hs (even if initially that involves putting some
of the gtk2hs syntax quirks into mainline c2hs while gtk2hs migrates to
the standard syntax -- or new agreed syntax).


More information about the cabal-devel mailing list