We need to have a header which claims the foreign imports we do are
for entities that actually exist. Otherwise the C compiler complains.

It wasn't long enough on thorkil's Mac. Should really do something based
on the calibration value or something, I guess.

The test is still expected to fail, as there is some 
implementation work to be done here

The script was using Data.Set to test handling of unboxed fields by :print,
but some change in Data.Set or in ghc has made fromList more strict and 
the :print pprinter never gets to show the unboxed fields itself.
Moreover, print022 is a much better test for unboxed fields than this one

- Although the test passed when run by itself, it did sometimes fail
  with validate as the timing just didn't work out on a loaded machine.

- See also #1523

e.g. the /bin/sh on my Ubuntu 7.04 system

Fixes a problem where /dev/null doesn't exist on mingw.