Why does alanz's branch blow up GHC's heap?
alanz mentioned that moving all of GHC's
Data instances instances to a new module causes the stage1 compile to run out of memory during compilation (he has 8 GB of memory). This is rather suspicious as this module isn't terribly large. IIRC typechecking time may be non-linear in program size, but I don't see why it would be non-linear in space. Investigate.