On Wed, 2006-10-25 at 01:36 +0100, Duncan Coutts wrote:
I would like to re-propose the last scheme that I cam up with. I'll try to make it a concrete proposal as well as giving some motivating examples to give the intuition of the meaning.
As a quick addendum, some comments arising from discussion with Ian... We may decide to ban more stuff. For example Ian suggested that it might not be sensible to let packages conditionally change the exposed modules. His example was using a package on Windows and then trying to use in on Linux and finding the exposed modules were all diferent. Changing hidden modules presumably would be ok. On the other hand, in Gtk2Hs I know one case where we do this. We have a Graphics.UI.Gtk.Cairo api module that is only included if Gtk was built against Cairo. In any case it could be faked by using cpp to just not export anything rather than not having the module exposed at all. So it's not clear that it's worth banning. Or maybe making it slightly harder is worth it so that people don't get in the habit. Ian also pointed out that some tests were missing, like a compiler test. I presume he was thinking of something like: configuration: implementation(hugs) some-thing: This is good stuff to discuss, but I think really those are the easy bits and we should figure out the tricky bits first. Duncan