🚀 go-pugleaf

In article <1992Jul27.221133.27847@mintaka.lcs.mit.edu> mycroft@hal.gnu.ai.mit.edu (Charles Hannum) writes:
>
>> mknod testpipe p
>> cat testpipe &
>> ls > testpipe
>> The "cat" process ends after the ls is sent to the screen.

>This is the way named pipes have worked since The Dawn of Time.  If you
>want to have multiple processes write to the pipe, you need the reader
>to reopen it.  In the shell, you could do something like this:
>(while :;do cat testpipe;done)&

Cat just happens to be written to exit when it hits EOF.  So, either
prevent it from seeing EOF by holding the write side open with an
unused fd:
  cat testpipe 4>testpipe &
or use a process that doesn't exit at EOF.  Just be sure it sleeps
a while between reads when there is no input.  A shell script that
periodically reads a control fifo makes a handy server:
do
 read A B
 ..process control command (if any) ..
 ..do some work...
done < fifo

You just have to start it in the background and throw it a real or
dummy command command to unblock its open on the fifo, then you
can "echo command argument > fifo"  manually or from other programs
to control it.  (Try that with sockets...).  If you want it to
wait for each command, open for write also ( done <fifo 4>fifo) 
otherwise the reads are non-blocking after the fifo has been
opened and closed by a writer.

Les Mikesell
  les@chinet.chi.il.us