pipes? threadWaitRead?

[Simon Marlow]

> I have some GHC-compiled programs that run OK under Sparc/Solaris but
> hang under x86/Linux.  I'm using GHC 5.02.3 on both systems (although
> I've tried GHC 5.03.20020410 on Linux with the same failures).  I'm
> debugging to narrow the problem down.  My current suspicion is that
> closing the write end of a pipe is not always causing the read end to
> see end-of-file.  Is there any strangeness relating to 
> this--or perhaps
> with `threadWaitRead` on pipes--that anyone knows about?  I sometimes
> use a pipe to have two GHC threads communicate within a 
> single process.

Is it named pipes or anonymous pipes?  There is weirdness to do with
FIFOs caused by different OSs implementing different semantics.  I
wouldn't be surprised if there were differences in the behaviour of
unnamed pipes too.

See the FAQ and previous msgs on this list for info on the FIFO
problems.

Cheers,
	Simon