Re: Detect whether attached or detached from screen session?

[Peder Stray]

On Tue, 14 Oct 2008, Mark Eichin wrote:

> Peder Stray <address@hidden> writes:
>
> > On Tue, 14 Oct 2008, Trent W. Buck wrote:
> >
> > > "Sven Hazejager"
> > > <address@hidden> writes:
> > >
> > > > I'm using screen to keep some applications open over a remote SSH
> > > > connection that sometimes is killed. What I would like to do, is to
> > > > somehow find out whether I am logged in (attached) or disconnected
> > > > from the screen session.
> > > >
> > > > Reason is that I want procmail to send incoming mail to my BlackBerry
> > > > if I'm disconnected, but NOT do that when I'm SSH-ing into my Pine
> > > > session.
> > >
> > > I don't know how to achieve that, but perhaps there's a different
> > > approach to the problem -- such as always forwarding all mail to an
> > > IMAP server that you can access from both the blackberry and from
> > > pine.
> >
> > The screen-socket has execute permissions when attached, and none when
> > detached, like this:
> >
> > attached:
> >
> > ls -l /tmp/uscreens/S-peder/4598.irc
> > prwx------ 1 peder peder 0 Oct 14 00:00 /tmp/uscreens/S-peder/4598.irc
> >
> > detached:
> > ls -l /tmp/uscreens/S-peder/4598.irc
> > prw------- 1 peder peder 0 Oct 14 07:06 /tmp/uscreens/S-peder/4598.irc
> >
> > Guess you can test for that i procmail by something like:
> >
> > :0c
> > *? perl -e 'exit ! grep { (stat $_)[2] & 0100 } </tmp/uscreens/S-peder/*>'
> > address@hidden
>
> Parsing "screen -ls" output (for "Attached" or not) might be less
> sensitive to implementation details, but that's a pretty clever test...
>
> I'm not familiar with procmail, but it might be cleaner to use
>
>  ! test -x /tmp/uscreens/S-peder/*
>
> (to avoid starting perl on every message - I think both of these assume 
> there's only one socket that matches the rule, though it's pretty easy 
> to change if not...)

True, perl is bigger than test, but i guess you mailserver should be real 
slow for you to actually notice :)  The perl test actually checks for any 
attaced screen, so it should work for cases where there is more than one 
socket in the dir.  Using the test-line as above would break violently in 
that case since it expect only one argument.  Parsing the output from 
screen -ls would of course be even more expensive that either perl or 
test.

Of course, all of this expects that screens tmp-dir is available on the 
mailserver, which it may not be.  In that case, it may be easier to make a 
wrapper for screen that touches a statusfile before starting screen and 
maybe doing the perl-test after and remove the file if no screens are 
attached (so you actually can attached more than once to the screen)

--
  Peder Stray