bug report

[Ed L Cashin]

Markus Friedl <markus.friedl at informatik.uni-erlangen.de> writes:

> On Wed, May 23, 2001 at 10:49:54PM -0400, mugz wrote:
> > 
> > I'm always a bit slow to report bugs i see, figuing someone else will
> > report it and that it will eventually get fixed. This one has been
> > somewhat of a problem for a while now. I run Linux Slackware -current and
> > just upgraded to OpenSSH 2.9p1, but I have noticed this same bug on every
> > platform and OS running OpenSSH.
> > 
> > If a user logs in and starts any kind of background process such as
> > "screen" "bnc" "mud" "eggdrop" etc, the console "hangs" when they log out.
> > When I ssh to a shell running OpenSSH from work or from home, start a
> > screen or a bnc and then exit with that process running, I have to go to
> > another console and manually kill my connection locally. If i then log
> > back into the same shell and type: last -5 myuserid, it shows me as
> > "still logged in" from that previous session. You can test this easily by
> > logging in using ssh and typing:
> > 
> > sleep 200 &
> > exit
> 
> does
> 	sleep 1234 < /dev/null > /dev/null 2>&
> 	exit
> work?

You can make init the parent of the remote shell executing mud or
whatever.  This uses a subshell to launch the command (parentheses do
this in Bourne-type shell like bash), so when the subshell exits, init
inherits the sleep process:

  here$ ssh there
  there$ (sleep 1234 &)
  there$ exit
  here$

... leaving you free to logout without anything blocking.

-- 
--Ed Cashin                   PGP public key:
  ecashin at terry.uga.edu       http://www.terry.uga.edu/~ecashin/pgp/