187274 - scrolling selects can cause CPU intensive freeze (GTK theme issue)

Status: RESOLVED WORKSFORME

(Core :: XUL, defect)

Description

[R.K.Aa.]

linux trunk 2002123108 bug been seeing this for 3-4-weeks now

load https://bugzilla.redhat.com/bugzilla/query.cgi
grab the thumb on the "Components" select box
drag to bottom, then up to top - repeat untill freeze

That's just a sample of how i can reproduce this 100%
It's been an annoying bug for a while and usually happens when i try to set
search criteria in mozilla.org's bugzilla. I have to "killall mozilla-bin" when
the freeze occurs. Timer related? I'm using a P4/1.8G CPU

Comment 1

[R.K.Aa.]

if i ctrl+c break a trunk CVS build it in gdb after the freeze, and do a where,
i see this:

(gdb) bt
#0  0x42028909 in sigsuspend () from /lib/i686/libc.so.6
#1  0x4020c238 in __pthread_wait_for_restart_signal () from
/lib/i686/libpthread.so.0
#2  0x4020905b in pthread_cond_wait () from /lib/i686/libpthread.so.0
#3  0x401db9a5 in PR_WaitCondVar () from
/home/dark/MOZ/TREE1/mozilla/dist/bin/libnspr4.so
#4  0x401715eb in nsThreadPool::GetRequest(nsIThread*) ()
   from /home/dark/MOZ/TREE1/mozilla/dist/bin/libxpcom.so
#5  0x40171e32 in nsThreadPoolRunnable::Run() () from
/home/dark/MOZ/TREE1/mozilla/dist/bin/libxpcom.so
#6  0x40170695 in nsThread::Main(void*) () from
/home/dark/MOZ/TREE1/mozilla/dist/bin/libxpcom.so
#7  0x401e0545 in _pt_root () from /home/dark/MOZ/TREE1/mozilla/dist/bin/libnspr4.so
#8  0x4020a981 in pthread_start_thread () from /lib/i686/libpthread.so.0
#9  0x4020aa85 in pthread_start_thread_event () from /lib/i686/libpthread.so.0
(gdb)

I have a "feeling" the freeze is to some degree avoidable if i drag the thumb
very very slowly.

Comment 2

[Brant Gurganus]

Can't reproduce on Windows with or without JavaScript.  The JavaScript version
takes forever to load.

Comment 3

[Andrew Schultz]

worksforme with linux trunk 20030101, scrolling slow or fast
the stack in comment 1 appears to be from a secondary thread.  what does the
stack from the main thread look like?

Comment 4

[R.K.Aa.]

That is the main thread. I'm running on RH8, if that matters.

Comment 5

[R.K.Aa.]

another thing: I wonder if this is some weirdness in interaction with GTK widgets:
when Mozilla freeze i can sometimes still move the scrollbar thumb of the select
box.

Comment 6

[R.K.Aa.]

I changed GTK theme from one called "Pretty Okayish Preon" to the simpler
"Pretty Okayish" - and can't reproduce the freeze anymore. WFM for now.

Comment 7

[R.K.Aa.]

still happening - reopen. Stack is weird but the same stack is also in bug 187966

Comment 8

[R.K.Aa.]

changing component. I still see this bug, and it pretty much prevents me from
doing any meaningfull searches in bugzilla.

Comment 9

[John Morrison]

rkaa: so the change in gtk theme in comment 6 wasn't truly WFM?

Is this with GTK 1 or GTK2? What is the output of 

/usr/bin/gtk-config --version
/usr/bin/gtk-config-2.0 --version

Comment 10

[R.K.Aa.]

I have both GTK1 and 2 installed. 
/usr/bin/gtk-config --version
1.2.10

/usr/bin/gtk-config-2.0 --version
bash: /usr/bin/gtk-config-2.0: No such file or directory

$ rpm -qa gtk*
gtkhtml2-2.0.1-2
gtk2-engines-1.9.0-4
gtk+-devel-1.2.10-22
gtk2-devel-2.0.6-8
gtkglarea-1.2.2-13
gtk-doc-0.9.0.200209080026-0.snap.ximian.5
gtk+-1.2.10-22
gtk2-2.0.6-8

Comment 11

[Andrew Schultz]

did you compile with gtk2 support?

Comment 12

[R.K.Aa.]

It isn't so interesting what i compile with, since i see the bug in official
builds as well. (But no, i don't compile with gtk2 support)

Comment 13

[Brian Ryner (not reading)]

I'm almost sure that isn't the stack from the main thread.  Type "thread 0"
before getting the backtrace.

Comment 14

[R.K.Aa.]

When the freeze occure, i can still scroll the widget, but it only scrolls down
regardless of which way i scroll it. When at bottom, that too is dead.
So - i do ctrl+c and..

Program received signal SIGINT, Interrupt.
[Switching to Thread 32773 (LWP 26817)]
0x420abaf1 in nanosleep () from /lib/i686/libc.so.6
(gdb) thread 0
Thread ID 0 not known.
(gdb) bt
#0  0x420abaf1 in nanosleep () from /lib/i686/libc.so.6
#1  0x400dc358 in __pthread_timedsuspend_new () from /lib/i686/libpthread.so.0
#2  0x400d94b9 in pthread_cond_timedwait_relative () from /lib/i686/libpthread.so.0
#3  0x400ab714 in pt_TimedWait () from
/home/dark/MOZ/TREE1/mozilla/dist/bin/libnspr4.so
#4  0x400ab8d2 in PR_WaitCondVar () from
/home/dark/MOZ/TREE1/mozilla/dist/bin/libnspr4.so
#5  0x40548121 in TimerThread::Run() ()
   from /home/dark/MOZ/TREE1/mozilla/dist/bin/libxpcom.so
#6  0x405448d9 in nsThread::Main(void*) ()
   from /home/dark/MOZ/TREE1/mozilla/dist/bin/libxpcom.so
#7  0x400b04ed in _pt_root () from /home/dark/MOZ/TREE1/mozilla/dist/bin/libnspr4.so
#8  0x400da981 in pthread_start_thread () from /lib/i686/libpthread.so.0
#9  0x400daa85 in pthread_start_thread_event () from /lib/i686/libpthread.so.0
(gdb) quit
The program is running.  Exit anyway? (y or n) y
[dark@localhost bin]$

Comment 15

[Brian Ryner (not reading)]

In that case, can you also give the output of "info threads" in gdb?

Comment 16

[R.K.Aa.]

(gdb) info threads
* 6 Thread 32773 (LWP 27139)  0x420abaf1 in nanosleep () from /lib/i686/libc.so.6
  4 Thread 16387 (LWP 27137)  0x42028909 in sigsuspend () from /lib/i686/libc.so.6
  3 Thread 8194 (LWP 27136)  0x420cfaeb in poll () from /lib/i686/libc.so.6
  2 Thread 16385 (LWP 27135)  0x420cfaeb in poll () from /lib/i686/libc.so.6
  1 Thread 8192 (LWP 27133)  0x4029082e in gdk_window_get_parent ()
   from /usr/lib/libgdk-1.2.so.0

   (gdb) thread 1
[Switching to thread 1 (Thread 8192 (LWP 27133))]#0  0x4029082e in
gdk_window_get_parent
    () from /usr/lib/libgdk-1.2.so.0
(gdb) bt
#0  0x4029082e in gdk_window_get_parent () from /usr/lib/libgdk-1.2.so.0
#1  0x40e2692a in gdk_window_child_of_gdk_window(_GdkWindow*, _GdkWindow*) ()
   from /home/dark/MOZ/TREE1/mozilla/dist/bin/components/libwidget_gtk.so
#2  0x40e266dc in handle_gdk_event(_GdkEvent*, void*) ()
   from /home/dark/MOZ/TREE1/mozilla/dist/bin/components/libwidget_gtk.so
#3  0x4027b2d5 in gdk_event_dispatch () from /usr/lib/libgdk-1.2.so.0
#4  0x402b097e in g_main_dispatch () from /usr/lib/libglib-1.2.so.0
#5  0x402b0e59 in g_main_iterate () from /usr/lib/libglib-1.2.so.0
#6  0x402b10f4 in g_main_run () from /usr/lib/libglib-1.2.so.0
#7  0x401af6df in gtk_main () from /usr/lib/libgtk-1.2.so.0
#8  0x40e1da38 in nsAppShell::Run() ()
   from /home/dark/MOZ/TREE1/mozilla/dist/bin/components/libwidget_gtk.so
#9  0x40e00cee in nsAppShellService::Run() ()
   from /home/dark/MOZ/TREE1/mozilla/dist/bin/components/libnsappshell.so
#10 0x08058382 in main1(int, char**, nsISupports*) ()
#11 0x08058a89 in main ()
#12 0x420159b4 in __libc_start_main () from /lib/i686/libc.so.6

Comment 17

[Brian Ryner (not reading)]

This doesn't sound like a gtk nsITheme issue to me, from the stack.  Blizzard,
any ideas?

Comment 18

[Christopher Blizzard (:blizzard)]

No, no idea.  I certainly don't see it, either.

Comment 19

[Ben Hsu]

Hello
I am hitting a bug which matches the description for this one. It reproduces on
the following releases: 1.0.2, 1.2.1, 1.3.1b

To reproduce, go to http://bugzilla.gnome.org/query.cgi, go to each multi-select
box with a scroll bar, and use shift-click to select all items. After 1-3 such
actions mozilla will hang and top will show CPU usage > 97%

Comment 20

[Andrew Schultz]

Ben: that's bug 137867, and is unrelated to this one.

Comment 21

[R.K.Aa.]

Haven't seen this one for a long time - something is fixed - somewhere. WFM.