Hi Claus,
Am Fri, 23 Jan 2009 07:55:05 -0500 schrieb "Lund, Claus" Claus.Lund@state.vt.us:
I checked out your previous mails... And I don't know if it's related. I haven't dived into the code or done any extensive testing to try to figure out why Stunnel was failing when using "exec = ...". And I haven't had any problems when using it the way we do.
the "exec" issue is not the point, that was problematic for me. The problem is the still open stdout/stderr file handle. See my mail from the 10th of January for details.
Could you try to run stunnel from a terminal instead?
I can (and did yesterday when I was helping Tom) ... but I'm not entirely sure what you'd like me to test :-)
I did the following:
1) running stunnel 4.20 (no separate libwrap): The stunnel processes are properly disconnected from stdout/stderr. See the output of "ps" - especially the "tty" column.
2) running stunnel 4.21 (with libwrap): The stunnel processes (except for one) are still connected to stdout/stderr. This can cause programs to hang (e.g. the cryptonas python code, that I referred to, was waiting forever for the daemon process to finish/disconnect - with 4.20 it worked).
3) running stunnel 4.21 (disabled libwrap): The stunnel processes are disconnected from stdout/stderr. It behaves just like 4.20 - no hang or other problems.
If you notice above behaviour (2) for 4.21 (or later) with libwrap enabled, then I guess, this is a problem, or? At least, I expected stunnel to disconnect completely from the terminal when entering daemon mode. If my expectations are flawed here, then we all can safely ignore this non-issue :)
thanks for your time! Lars