[1440] in Athena Bugs
xterm -C and its implications for lingering processes
daemon@ATHENA.MIT.EDU (qjb@ATHENA.MIT.EDU)
Tue Nov 29 13:00:38 1988
From: <qjb@ATHENA.MIT.EDU>
Date: Tue, 29 Nov 88 12:59:51 EST
To: bugs@ATHENA.MIT.EDU
I have noticed a bug on RT's running 6.0R which starts the login xterm
with the -C option. I have not seen this on a vax, but I haven't used
a vaxstation since 6.0R came out!!!
I haven't figured out how to reliably produce this bug, but I am
looking. All I can include here is a rough idea of what is going on.
Maybe this will give someone the necessary lead. I'll write a
detailed bug report if/when I figure out what exactly what is going on
and how to reliably reproduce the problem.
When a process that has /dev/ttyv0 open (or whatever device is getting
console output), does not terminate when the X server dies, the
console output stays pointed to /dev/ttyv0 (or whatever terminology
you would use to refer to this) rather than the console. This means
that you don't see anything on the console. The result in a public
cluster is that there is a workstation with a screen containing only a
cursor. This workstation has a load average of around 1 because there
is a process running that hasn't released /dev/ttyv0. The fix is to
activate the workstation (it will activate eventually if you keep
hitting keys - I haven't quite figured this out either - and it's not
that it activates on the first key and you don't see the output, the
disk noise doesn't start right away at all) and kill the process. You
can't do a control-P login (or I have not succeeded in doing so.) I
tried, and couldn't do it, even without being able to see; i.e., I got
no evidence that I was actually logged in...
Hopefully more and better info to come, unless someone knows right
away what to do.
