Daily report for head

[Claus Reinke]

> FWIW, I think we could do a lot to improve the reporting of test failures in the buildbot status 
> emails.  I usually tend to look at the actual logs.

one thing that confused me with the  macgyver PPC OSX head logs

http://darcs.haskell.org/buildbot/head/macgyver%20PPC%20OSX%20head/builds/23

was that i couldn't understand how the test even got this far. the
'unexpected' log reports the failure of ':set -package ghc' in ghci024,
but the stdout preparation for the test (prepare024) already does a
'ghci -package ghc', and that earlier error was nowhere to be seen.

in the full 'stdio' log, that error appears just before the test header:

ghc-6.9.20071107: internal error: unconditional relative branch out of range: jump island out of 
range
    (GHC version 6.9.20071107 for powerpc_apple_darwin)
    Please report this as a GHC bug:  http://www.haskell.org/ghc/reportabug
=====> ghci024(ghci)

shouldn't the setup code be run after outputting the test header?
or is it a case of stderr being faster than stdout?

claus