For those who asked, here are excerpts from emails I've sent
to Debian regarding the CTWM crashes that I've repaired.
#################################################################
A brief sketch of typical usage and failure mode:
I am running CTWM with eight workspaces (most are populated)
and switching workspaces fairly often. The most active
X apps are xterm, Firefox, exmh and mahjongg. I typically
have several Firefox windows open, each of which has multiple
tabs active, sometimes dozens. Failure seemed most often
to happen immediately after I switched workspaces to one
containing a Firefox window with a *LOT* of tabs open.
I've not seen it even once since this I made this fix.
Note that I was usually able to recover my entire X session
and continue without harm if I could manage to launch a
new instance of CTWM from the console or an SSH session.
Until I patched my CTWM I actually had a shell loop right
inside my .xsession file that would automatically relaunch
CTWM when it crashed, so I was able to live with it for
several years before finally getting fed up and fixing it!
#################################################################
Here's a bit of GDB output to illustrate the specific
fauilt that was constantly occurring before I fixed it:
Program received signal SIGSEGV, Segmentation fault.
0x0805f131 in PlaceOntop (ontop=8, where=0) at menus.c:3756
3756 for (t = Scr->TwmRoot.next; t != NULL; t = t->next) {
(gdb) where
#0 0x0805f131 in PlaceOntop (ontop=8, where=0) at menus.c:3756
#1 0x0805f24b in RaiseWindow (tmp_win=0x83343a0) at menus.c:3787
#2 0x0806ab10 in AutoRaiseWindow (tmp=0x83343a0) at events.c:218
#3 0x0807096f in HandleEvents () at events.c:547
#4 0x0804e3a7 in main (argc=2, argv=0xbff491a4, environ=0xbff491b0)
at ctwm.c:1027
(gdb) print Scr
$2 = (ScreenInfo *) 0x0
#################################################################
To: Debian Bug Tracking System <[email protected]>
Date: Tue, 06 Mar 2012 15:42:25 -0500
Subject: debianBug#662860 - ctwm: SIGSEGV when changing workspaces
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=662860
Package: ctwm
Version: 3.7-3.2
Severity: important
Tags: upstream patch
Dear Maintainer,
I'm a longtime CTWM user who finally got tired of the random
crashes that occur, typically when I'm switching workspaces.
I can sometimes go for a month without seeing them and sometimes
it happens several times a day. I've analyzed this to the point
where I can see that code in DispatchEvent() is occasionally
leaving the global Scr pointer NULL which causes a SIGSEGV on
our next pass through the loop in HandleEvents().
I've provided a patch that adds a few lines of defensive code
which seems to have improved the situation, though I have not
analyzed the root cause, which appears to be that, for unknown
reasons, either XFindContext() or FindScreenInfo() seem to be
unable to return the answer expected of them.
--Michael O'Donnell
BEGIN PATCH:
--- events.c.b0rken 2012-03-06 15:10:42.093677349 -0500
+++ events.c 2012-03-06 15:19:52.369161662 -0500
@@ -435,21 +435,25 @@
*/
Bool DispatchEvent2 (void)
{
Window w = Event.xany.window;
StashEventTime (&Event);
+ ScreenInfo *lastScr = Scr; /* XXX_MIKE - assume Scr OK on entry... */
if (XFindContext (dpy, w, TwmContext, (XPointer *) &Tmp_win) == XCNOENT)
Tmp_win = NULL;
if (XFindContext (dpy, w, ScreenContext, (XPointer *)&Scr) == XCNOENT) {
Scr = FindScreenInfo (WindowOfEvent (&Event));
}
dumpevent(&Event);
- if (!Scr) return False;
+ if (!Scr) {
+ Scr = lastScr; /* XXX_MIKE - try not to leave Scr NULL */
+ return False;
+ }
if (Scr->Root != Scr->RealRoot) FixRootEvent (&Event);
#ifdef SOUNDS
play_sound(Event.type);
#endif
@@ -473,21 +477,25 @@
*/
Bool DispatchEvent (void)
{
Window w = Event.xany.window;
StashEventTime (&Event);
+ ScreenInfo *lastScr = Scr; /* XXX_MIKE - assume Scr OK on entry... */
if (XFindContext (dpy, w, TwmContext, (XPointer *) &Tmp_win) == XCNOENT)
Tmp_win = NULL;
if (XFindContext (dpy, w, ScreenContext, (XPointer *)&Scr) == XCNOENT) {
Scr = FindScreenInfo (WindowOfEvent (&Event));
}
dumpevent(&Event);
- if (!Scr) return False;
+ if (!Scr) {
+ Scr = lastScr; /* XXX_MIKE - try not to leave Scr NULL */
+ return False;
+ }
if (captive) {
if ((Event.type == ConfigureNotify) && (Event.xconfigure.window ==
Scr->CaptiveRoot)) {
ConfigureRootWindow (&Event);
return (False);
END PATCH
-- System Information:
Debian Release: wheezy/sid
APT prefers testing
APT policy: (990, 'testing'), (500, 'unstable')
Architecture: i386 (i686)
Kernel: Linux 3.1.0 (SMP w/2 CPU cores)
Locale: LANG=en_US, LC_CTYPE=en_US (charmap=ISO-8859-1)
Shell: /bin/sh linked to /bin/bash
Versions of packages ctwm depends on:
ii libc6 2.13-26
ii libice6 2:1.0.7-2
ii libjpeg8 8d-1
ii librplay3 3.3.2-14
ii libsm6 2:1.2.0-2
ii libx11-6 2:1.4.4-4
ii libxext6 2:1.3.0-3
ii libxmu6 2:1.1.0-3
ii libxpm4 1:3.5.9-4
ii libxt6 1:1.1.1-2
ii m4 1.4.16-2
ii x11-common 1:7.6+11
ctwm recommends no packages.
ctwm suggests no packages.
-- Configuration Files:
/etc/X11/ctwm/system.ctwmrc [Errno 2] No such file or directory:
u'/etc/X11/ctwm/system.ctwmrc'
-- no debconf information
-- debsums errors found:
debsums: missing file /usr/bin/ctwm (from ctwm package)
Note that the above mentioned debsums error for
/usr/bin/ctwm is due to my having built and installed
a private version from source during debug....
