[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: QuickCheck version

Ouch, this is what I get for being overeager to update libraries.
Parsec and now this, too.  I'll revert Parsec to 2.x if that's
necessary.  Sorry about this.

On Wed, Mar 11, 2009 at 09:42:47AM -0500, John Goerzen wrote:
> I would advocate replacing libghc6-quickcheck-dev with the 1.x version
> to match upstrea, and adding a new libghc6-quickcheck2-dev for those
> that need 2.x.  There are significant API changes.

I'd rather rename the current 2.x package as libghc6-quickcheck2-dev
and add libghc6-quickcheck1-dev and have both provide
libghc6-quickcheck-dev.  1.x may still be the preferred one but 2.x is
where the new stuff happens, presumably.

Speaking of multiple versions, currently, haskell-devscripts'
generated postinst scripts hide any other versions of packages with
the same cabal name.  Is this the right thing to do?  Having which one
is exposed decided by the order of postinst calls sounds uncomfortably
undeterministic.  If there are multiple visible versions, then Cabal
can pick the correct version during build if the right constraints are
in place.  We'd need to put build-conflicts in place for any other
versions of a package or make them conflict with other versions.
Neither seems that appealing.

Was the problem case when a library would pull different versions of a
single library via its deps?  I'm not sure if that's so much of an
issue in Debian.  Hiding doesn't fix packages that'd need to be
rebuilt against a newer version and if there are multiple versions of
libraries included in Debian with incompatible APIs then the libraries
using those are supposed to use a constraint to pick one.

Reply to: