Hey,

I am experimenting a client server model using 2 name pipes. I want the client to accept input from terminal and then send to the server, and also accept response from server and then send to the screen. The server is basically "bash -i".

I am trying this:

server$ mknod pipe1 p; mknod pipe2 p
server$ bash -i <pipe1 >pipe2 2>&1

client$ (in another terminal)
while true; do
read a
echo $a > pipe1
cat pipe2
done;


The problem is that I can just do the command once and then the server exit by itself. I wonder if there is any ctrl-d character wrongly sent to the server and hence closing it.

Would somebody please help? Thanks!

echo $a > anyfile would delete anyfile first (or maybe just empty it?) but I don't know what effect it would have on a named pipe. However, I would try echo $a >> pipe1 and see if this makes a difference.

Thank you very much for your suggestion. I just tried ">>" and also trying changing the shell script to use "#!/bin/bash" instead of "#!/bin/sh", but it didn't seem to make a difference.

Hey,

Now I am guessing the problem was this: if I use
echo $a > pipe1
cat pipe2

The named pipe will be closed once the command is completed, and hence close the server as well (because the server [bash -i] relies on these named pipes to be the input and output.

Therefore, I am trying this now:


#!/bin/sh
exec 5>pipe1
exec 6<pipe2
while true; do
read a
echo $a >&5
cat <&6
done;

I am trying to keep the named pipes open by using "exec 5>pipe1" and "exec 6<pipe2", but this time I got a deadlock in the line "cat <&6" after receiving some output from the server.

It seems that keeping the pipes open is not a good idea.

Actually, I tried using 2 terminals, with one running "cat pipe2" and one running "cat >pipe1" and I succeeded. However, what I actually want to do is to have a client remotely controlling the server as if working on the server.

Any idea?