GNOME Bugzilla – Bug 169684
Certain SDL Apps kill Metacity (problems with 32 bit visuals and/or themes?)
Last modified: 2016-02-07 23:21:07 UTC
Steps to reproduce: 1. Run an SDL app-- Ones that I've had do this are: ) Tecnoballz ( http://linux.tlk.fr/games/TecnoballZ/ ) ) Fceu ) abuse-sdl (not sure where to obtain these two, I used the Ubuntu packages) 2. Switch to fullscreen mode (f on trackballs, alt-enter on fceu, --fullscreen with abuse) 3. Metacity crashes-- This is especially bad because, due to mouse cursor locks, it may require a restart of the X server to get mouse functionality back. Stack trace: I wasn't able to get a useful stack trace-- according to GDB, metacity sent SIGABRT instead of SIGSEGV, but I did get this message on stdout, when the SDL window got focus. Bug in window manager: Unexpected X error: BadMatch (invalid parameter attributes) serial 2790 error_code 8 request_code 70 minor_code 0) Aborted Other information:
Looks similar to the error that Fernando reported to me that he got a little while ago and could get repeatedly but for some obscure webcam app so I couldn't investigate further. I think we also have another bug in here that may be the same--obscure BadMatch X error. Anyway, could you get us a verbose debugging log? To do so: - Reduce your desktop to as few windows as possible to reproduce the bug - Run METACITY_VERBOSE=1 METACITY_USE_LOGFILE=1 metacity --replace - On stdout metacity will print the name of the logfile - Reproduce the bug as quickly as possible - Kill the metacity you started above to stop the logfile from growing any longer - Attach the logfile here (you may have to compress it first, though hopefully you won't have to)
Created attachment 38469 [details] Log of the a crash
A little research reveals this could be theme-related somehow-- the crash happens when I use the Atlanta theme (or my custom theme, loosely based on Atlanta), but not when using the Metabox theme.
A backtrace with --sync or METACITY_SYNC=1 would show exactly what causes the BadMatch
Unfortunately, that didn't yield a better stack trace-- it looked identical to the one I had without METACITY_SYNC set-- a lot of "0x00000000 in ??" lines, then a single libc call and a message about possible corrupt stack. Is the crash log I posted earlier adequate to track down the problem? If not, perhaps I need to do something else to obtain a workable trace.
sounds like a buffer overflow
It's not a buffer overflow, we die with SIGABRT on X error. I would say either no debug symbols or standard gdb flakiness.
Probably related to bug 171174.
Found a possible lead on this. I hacked SDL to not ask for 32bit visuals, to stop them from looking weird when xcompmgr is loaded. (unrelated to this bug), and I couldn't reproduce this bug's crashes at all.
Possibly also related to bug 314442? Updating the title and hoping that the 32 bit visuals hint in comment 9 (and perhaps the theme hint in comment 3) is helpful to someone that has a clue about what visuals are and how they are used...
Reporter: has this occurred for you since 2005?
I haven't used metacity for a while, but when i tried it just now, it works fine. I couldn't say when it started working.