On Wed, 2009-09-30 at 23:13 +0200, Andreas Schwab wrote: > > Just make sure the write side of the pipe is not closed prematurely.
Hrm. Yes, of course. John's solution of having a null writer keeping it open is one way -- which I might just use. > $ (n=0; while [ $n -lt 10 ]; do cat /dev/zero; let n=$n+1; done) > /tmp/fifo & But this is where (simplified) my example using cat went sideways. :-( In my real world use, the first cat is actually mplayer and doesn't have the option of writing to stdout instead of a named file for this particular use of it. So other than John's solution, I suppose I could still follow your advice of not having the fifo write close prematurely and insulate the writer from the fifo. I can't think of another way other than using two fifos with a cat between them. Can anyone else? FWIW, the actual process which the second cat is standing in for is oggenc and it can take stdin, if that makes it any easier. So to recap: mplayer ... /tmp/fifo & oggenc ... /tmp/fifo The mplayer may terminate prematurely and need to be restarted, ideally without oggenc knowing this. Cheers and much thanx! b.
signature.asc
Description: This is a digitally signed message part
