On Thu, 2009-11-12 at 17:37 +0000, Sam Martin wrote: > Although it might be a pain in the arse to some degree, is there any > reason why 'base' is considered special? > > As an example, I've come across a fair number of libraries/apps that > (presumably) compile against a previous version of OpenGL, but not the > current latest.
The plan eventually is to do this for all packages that opt-in to following the PVP[1]. Base follows the PVP and it's a bit special since it's the single package that causes most breakage when new major versions come out. So we're dealing with part of the problem now as a special case and the rest of the problem later when we've got the appropriate infrastructure. > Given it's impossible to test any package against libraries that don't > yet exist, shouldn't the upper bound be required for all package > dependencies? True, however when you write that dependency you have no idea what that upper bound ought to be unless you know the package is following some kind of version policy. That's why we would only enforce it for packages that opt-in to the PVP. [1]: http://www.haskell.org/haskellwiki/Package_versioning_policy Duncan _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
