On Tue, 2009-10-13 at 08:48 -0500, John Goerzen wrote:
Now I'm getting complaints from people using 6.10.4 saying that there are now missing instances of Typeable with time 1.1.2.4.
Right, because 1.1.2.4 is an earlier version than 1.1.3 which is the random intermediate snapshot version included in the ghc-6.10.3 extralibs collection and the version where the Typable instances were added. So the problem you're noticing is that some people are using 1.1.2.4 and others are using 1.1.3 or 1.1.4 and these are not related to the version of ghc that they are using, so using the ghc version as a proxy fails.
1) Did the Typeable instances get dropped again from time?
No.
2) What exactly should I do so this library compiles on GHC 6.8 and 6.10.x?
Depend on the time library and use one of the following techniques: If you're prepared to depend on Cabal-1.6 then you can use the cpp macros that let you do conditional compilation on the version of a package you depend on. You mention that the time only incremented the 4th digit when it added the instances but I don't think that's right. My time-1.1.2.4 has no Typeable instances but 1.1.4 does (and I believe 1.1.3 did too). So you should be able to use this mechanism. Alternatively you can use the "flag hack" in the .cabal file: flag annoying-time-instances library ... if flag(annoying-time-instances) build-depends: time >= 1.1.3 else build-depends: time < 1.1.3 cpp-options: -DUSE_OWN_TIME_TYPABLE_INSTANCES
so it appears that what's happening here is that GHC 6.10.3 extralibs included time 1.1.3, but then haskell-platform standardized on 1.1.2.4. This is pretty annoying -- that haskell-platform would standardize on a version older than what shipped with a GHC release -- but I guess I can work around it by restricting my build-dep to be time < 1.1.3 and re-adding the instances.
No, it's the haskell-platform that was and is doing the right thing and it is ghc that subsequently accidentally shipped a random development snapshot. The platform has a commitment to provide API compatible versions of packages within a major series. The first release of the platform used time-1.1.2.4 (along with ghc-6.10.2) and thus it could not include the time-1.1.3 that the subsequent release of ghc-6.10.3 accidentally included. Duncan