Hi,
I'm looking for a way to execute only one Door32 instance at the time, only one node can run it. I see the method on the Wiki using semaphore but if the
door crash the semaphore will never be cleared and the Door will never be reachable again.
Thanks.

... If you eat yogurt you'll have lots of culture.

--- Mystic BBS v1.12 A48 (Linux/64)
* Origin: Gods69 BBS (21:3/220)

Hi Gods69,

Yeah, the wiki semaphore trick has exactly that problem: if the door
crashes, the file sticks around and the door is dead until someone
deletes it by hand.

On Linux there's a cleaner way: wrap the door with flock. The kernel
releases the lock as soon as the process dies, crash or not. No stale
lock left behind.

Example for a D3 menu command:

/usr/bin/flock -n /mystic/locks/mygame.lock /path/to/yourdoor %P

Create /mystic/locks/ once. The lock file itself can stay on disk; what
matters is the active hold. With -n, flock fails immediately if another
node already holds it, so the door never starts twice. You can then show
a short "busy, try later" message (e.g. with a menu ACS / !E check, or
by checking flock's exit code in a tiny wrapper).

Tiny wrapper if you want the busy message:

#!/bin/sh
LOCK=/mystic/locks/mygame.lock
DOOR=/path/to/yourdoor
if ! flock -n "$LOCK" "$DOOR" "$@"; then
echo "Sorry, that door is already in use. Try again later."
sleep 2
fi

Point D3 at the wrapper instead of the door binary. Same idea works for
DD doors too.

If you want to stay with the wiki semaphore method, a wrapper that
checks whether the door process is still alive and deletes a stale
semaphore first also works. flock is just more robust.

Cheers
Goose

--- Mystic BBS v1.12 A48 (Linux/64)
* Origin: Goosenet BBS (21:1/177)

On Linux there's a cleaner way: wrap the door with flock. The kernel releases the lock as soon as the process dies, crash or not. No stale
lock left behind.

Thanks. I will try it...

... Communism is like one big phone company. (Lenny Bruce)

--- Mystic BBS v1.12 A48 (Linux/64)
* Origin: Gods69 BBS (21:3/220)