Controlling kermit shell from within CC

Hi folks,

I'm in a difficult situation where I need to be able to control the kermit shell from within my C-C++ application.

I tried the standard fork()+dup2() tricks to connect pipes to the child's (kermit's) stdin-out-err to the parent. However, when controlling kermit this way, kermit requires that the pipe connected to stdin be closed before it'll actually respond to the given commands (although I can see in kermit's debug.log that it received the commands and processed them). This is unacceptable, as I need to dynamically generate and send numerous commands to kermit from my application.

I think the problem in the above-mentioned approach is kermit "knows" that it's parent is not really a tty. So, I am experimenting with forkpty() to launch kermit under a pseudo-terminal, but I'm having problems with that approach as well (we can discuss this more, if necessary).

One annoying requirement is that kermit must remain open between commands, as closing it down is far too slow b-c of delays releasingdev-ttyS0.

Also, kermit is my application of choice because: a) it's nicely scriptable b) it can wait for particular prompts from the remote host c) it saves me the trouble of figuring out how to open, condition, utilize, and close the line to the remote host over a serial line

If it matters, I only need to make this work on Fedora Core 3 (Linux 2.6.9).

Anyway, I was hoping someone could give some advice-direction here.

Thanks, Adam