Version: (using KDE KDE 3.3.2) Installed from: Gentoo Packages Compiler: gcc 3.3.2, sometimes with a cross-compiler to Alpha from x86 OS: Linux I'm using an Alpha EV5 with Gentoo Linux. Programs this time were compiled with gcc 3.3.2 with flags "-O2 -mieee -mcpu=ev5 -fomit-frame-pointer -g" Anything that generates a file listing crashes. So far, I've tried FTP and local, and Konqueror, Kate, KPDF, aRTs Builder, and more. Applications that don't use KDE, e.g. Valknut, don't crash. File listings include a view of files in Konqueror, and seeing a bunch of files in an open/save dialog. The affected KDE programs usually don't crash immediately, but they run for a while until something happens and they all crash at the same time. One thing that usually precipitates the crash is launching a separate program from Konqueror to deal with a remote file. Clicking "Save" in a Save File dialog also usually precipitates the crash. Often, just waiting while doing unrelated things gets the programs to crash, but that takes longer. Actually, launching a separate program from Konqueror to deal with a remote file gets that instance of Konqueror to crash whether or not it has generated a file listing. Programs that haven't dealt with file listings don't crash like this, e.g. Konqueror instances used only for web browsing. I'm not sure why, but the desktop doesn't crash, either. They all have the same backtrace, except for the LWP: Using host libthread_db library "/lib/libthread_db.so.1". [Thread debugging using libthread_db enabled] [New Thread 16384 (LWP 25313)] 0x00000200022b0360 in waitpid () from /lib/libpthread.so.0 #0 0x00000200022b0360 in waitpid () from /lib/libpthread.so.0 #1 0x0000020000c3cc78 in KCrash::defaultCrashHandler(int) (sig=11) at kcrash.cpp:246 #2 0x00000200022af428 in __pthread_sighandler () from /lib/libpthread.so.0 #3 0x0000020002581ce0 in sigset () from /lib/libc.so.6.1 #4 0x000002000257ee28 in __open_catalog () from /lib/libc.so.6.1 #5 0x000002000137d7b0 in QEventLoop::enterLoop() () from /usr/qt/3/lib/libqt-mt.so.3
how about you compile without omit-working-backtraces and perhaps even with -O0 then?
That was interesting. Backtraces now have little more information, but it's less inconsistent.
Created attachment 11218 [details] Backtrace from Kate sitting in the background. Test case 1: 1. Launch Konqueror via the Home icon. 2. Launch Kate, look at the file structure a bit using the sidebar but don't open anything. 3. Right-click on a file in Konqueror and select delete. 4. Click Delete in the verifying dialog box. 5. See a "Delete File(s) in Progress" box appear and not go away until I click "Cancel"
Created attachment 11219 [details] Backtrace from Konqueror trying to delete a file
Created attachment 11220 [details] Backtrace from Konqueror launching Noatun 1. Launch Konqueror 2. Browse filesystem until I find something to open that'll get it to crash. 3. After launching KEdit, find an MP3 file and click it. 4. Success! Noatun launches, but Konqueror crashes.
Created attachment 11221 [details] Backtrace from Konqueror innocently sitting in the background 1. Launch Konquerer with Home icon. For good measure, have 2 such instances running. 2. Launch some other program with the intention of getting Konqueror to crash. 3. After opening a file in KWord, and closing KWord, find KPresenter. 4. Open a PPT presentation in KPresenter. 5. Success! Both Konqueror instances and KPresenter crash. Oddly, only 1 Konqueror bug handler appears.
Created attachment 11222 [details] Incomplete backtrace from KPresenter crashing Unfortunately, I forgot to recompile KOffice with debugging, and my computer takes a while to recompile stuff.
Er, make that the backtraces look a little more inconsistent. And, the CFLAGS that I gave at the time were: -O0 -mieee -mcpu=ev5 -g
but still all your backtraces end in Qt and libc. I suggest you either debug this on your own or go with bugs.gentoo.org - I can't help you here. Not without an alpha machine.
I thought it would be clear by now that I'm not useful at debugging. It's not so clear that the Gentoo/Alpha team is swamped with other stuff like server-related security updates and GNU toolchain upgrades. I've had a bug report there for the better part of a year. http://bugs.gentoo.org/show_bug.cgi?id=61434 Note the lack of any involvement so far. Anyway, I tried compiling QT with no optimizations and debugging, and then tried the Kate + Konqueror + deleting exercise. Minimal change to the Kate backtrace. Because the "program that sits in the background and crashes" backtrace looks so similar, and because KDE is extremely slow compiled this way and I'm way past my bedtime, I'm not doing the other exercises yet. The Konqueror + Delete exercise leaves a Deleting file window open indefinitely, but it does delete one file. If you tell it to delete a single file, it'll delete the file. If you tell it to delete an entire directory, it'll delete one file from the directory. I'm not sure if this belongs here or in another bug report. Using host libthread_db library "/lib/libthread_db.so.1". [Thread debugging using libthread_db enabled] [New Thread 16384 (LWP 24333)] 0x00000200024ae360 in waitpid () from /lib/libpthread.so.0 #0 0x00000200024ae360 in waitpid () from /lib/libpthread.so.0 #1 0x0000020000dc3e68 in KCrash::defaultCrashHandler(int) (sig=11) at kcrash.cpp:246 #2 0x00000200024ad428 in __pthread_sighandler () from /lib/libpthread.so.0 #3 0x000002000277fce0 in sigset () from /lib/libc.so.6.1 #4 0x000002000277ce28 in __open_catalog () from /lib/libc.so.6.1 #5 0x000002000157b7b0 in QEventLoop::enterLoop() (this=) at kernel/qeventloop.cpp:198
The problem seems to have vanished in a manner that I don't understand.