Regressions when removing base3-compat

[Duncan Coutts]

On Wed, 2009-06-03 at 12:55 +0100, Ian Lynagh wrote:

> > For a smoother transition away from base 3 (rather than just dropping it
> > and breaking packages), we are just about to make hackage require upper
> > bounds on the version of base and we're about to release an update to
> > cabal-install so that it will select base 4 when packages specify that
> > they can work with either 3 or 4 (where as currently it picks base 3
> > when there is a choice between them).
> 
> That won't affect most of the packages that would break. It might mean
> some of the 36 get fixed, but the 125 already have a dep base<4 or
> similar.

Right, for those packages we need to start giving them deprecation
warnings. Anything getting built against base 3 should be given clear
warnings. Would it work to just add module DEPRECATED pragma(s) on key
modules exported from base 3? The message would have to make it clear
it's not the module as such but the version of the package.

> That sounds sensible. And the HP also presumably wants to include syb
> anyway.

Right. It gets a free entry due to having been a core lib.

Duncan