On 23/07/2020 04:27, haoniukun via Exim-users wrote:
> I actually happen to see the same question recently.
> One problem I found is that if I use exit_ctl to stop the process, this file can be removed without any problem.
> But if I use -d option to start the exim daemon from the command line to check the debug output and then use ctrl+c to kill the process, the file exim_daemon_notify cannot be removed properly.
> As a result, I cannot start exim process.
> And for me, if I manually remove exim_daemon_notify file, exim can start successfully for me.
> It took me really a long time to figure this out though.
> I cannot find any relevant information from the search engine.
> But you need to make sure you only have one instance of exim running.
> I’m using a Mac. But I believe it applies to freebsd as well since your error message is exactly the same as mine, as well as the exim version.
Interesting.
Presumably, since you say you can manually remove the
notifier socket via its filename, its presence can be
seen using netstat? How is the socket name displayed
there?
The socket is unlinked by the daemon, on shutdown,
for the "abstract socket" case (only). For the
non-abstract case, we rely on the daemon closing it
to result in its disappearance. Perhaps the BSDs
are behaving differently here... I can test on a
FreeBSD, but not on a Mac; lets hope that is
good enough...
The close/unlink is done in the daemon just before a fork
for removing a pidfile. This for should be
visible in debug output; can you verify the daemon
gets to that point? It's about the last thing it
does on closedown.
--
Cheers,
Jeremy
