problems with GHC.Conc.forkProcess

[Dean Herington]

I'm still having problems with GHC.Conc.forkProcess.  Sometimes some
threads other than the main thread that does the forkProcess carry
through into the newly forked process.

In case it might have any bearing, the evidence for my above claim is
that in my tests these "phantom" threads often suffer an exception like:

failed
Action: lazyRead
Handle: {loc=<file descriptor:
4>,type=semi-closed,binary=True,buffering=block (8192)}
Reason: Bad file descriptor
File: <file descriptor: 4>

I'm using GHC 5.04.3 (.tar.bz2 form) on Linux.