On Thu, May 28, 2015 at 3:23 PM, Joachim Breitner <[email protected]> wrote:
> Hi David, > > > Am Mittwoch, den 27.05.2015, 15:39 -0700 schrieb David Fox: > > On Wed, May 27, 2015 at 1:04 PM, Joachim Breitner <[email protected]> > > wrote: > > > > Also, it opens a new can of worm: Do we have to provide > > libghcjs-foo > > packages for all our Haskell libraries? For none? For a few? > > > > We build more and more libghcjs-* packages. I did the > > haskell-devscripts patches to support this. > > that’s great! My worry is not so much whether it is possible to package > libraries for GHCJS (it is, thanks to your work), but more the > complications it causes for packaging Haskell. For example, what if we > upgrade library foo to a new version, and that fails to work with GHCJS. > Do we promise to patch it then? Do we simply drop the libghcjs packages > (and all its reverse dependencies) again? How much will it make Haskell > migrations even harder? Yes, I see what you mean. My only thought on the subject would be to recommend separate source packages for the ghcjs libraries, with names different from the ghc packages, so that upgrades of the haskell libraries don't force upgrades of the ghcjs libraries. Also, I don't think haskell-devscripts can manage a single source package building ghc and ghcjs binary packages. The downside to this is when you are trying to have the two communicate, but I think this is a minor issue. And its pretty amazing how many things "just work" under ghcjs. Anyway, ghcjs library debs probably not a high priority, just trying not to create additional obstacles when the time comes.
