[221] in SIPB_Linux_Development
Console problem
daemon@ATHENA.MIT.EDU (System PRIVILEGED Account)
Wed Nov 24 12:34:54 1993
From: root@MIT.EDU (System PRIVILEGED Account)
Date: Wed, 24 Nov 93 12:33:55 -0500
To: linux-dev@MIT.EDU
I recompiled /etc/athena/console (in /usr/athena/src/src/dash/console)
with -g, producing a 1.2MB binary. This is currently installed in
/etc/athena/console, with the old 64K binary (compiled with -O2 in
place of -g) in /etc/athena/console.normal.
I reproduced the spinning console problem (not hard; I just logged out
as root) and attached the process using gdb. Here's what I get:
(gdb) file /etc/athena/console
Reading symbols from /etc/athena/console...done.
(gdb) attach 29405
Attaching program `/etc/athena/console', pid 29405
0x30e48 in write ()
(gdb) next
Current function has no line number information.
Single stepping until function exit.
0x117f7 in _XFlush ()
(gdb) back
#0 0x117f7 in _XFlush ()
#1 0x3 in ?? ()
(gdb) next
Current function has no line number information.
Single stepping until function exit.
Program received signal 13, Broken pipe
0x30e48 in write ()
Repeat ad nauseum.
Since console doesn't catch SIGPIPE, this looks to me like a bug in
XFlush() in our Xlib. This would also explain why zwgc displays the
same behavior when the server dies on it.
I don't think we have the source for our current version of Xlib
on-line; I may get a chance to pull them over and look at it later,
but I'm not terribly enthusiastic about tracking down ugly
signal-handing bugs, and I haven't slept in 26 hours, so it will wait.
I can't think of a workaround in the meantime, either.
Minor monitor nit: I wasn't able to adjust the VRC-16 to get rid of
the overlapping scan line near the bottom (you can see it easily with,
say, xsetroot -solid red).
--GBH
Who really wants AFS for this thing; then it'd be just like a Dec, but faster.
