GNOME Bugzilla – Bug 87414
Nautilus crash when invoked with '-q' option
Last modified: 2009-08-15 18:40:50 UTC
I have the 2nd July sources from CVS. The 'Nautilus to draw desktop' /must/ be turned on for the crash to happen. Open a terminal and type 'nautilus -q' (this command line switch option is supposed to invoke nautilus and quit instantly). All nautilus processes are killed and they respawn. Even before the nautilus comes up, type 'nautilus -q' again and again in quick succession. Nautilus crashes. Reproducibility is a little poor in Linux as compared to the Solaris platform. In Linux, most of the times, i get a "Nautilus cant be used now, due to an unexpected error" dialog. But it did crashed once in a while. In Solaris, the crash is very frequently observed. Sometimes, it occurs due to a bus error, sometimes a segfault and at time it aborts. But one of these three /definitely/ occur. Pasting the stack trace from Linux bug buddy. The Solaris stack traces are also very similar. Please feel free to ask for any specific traces, if required. I'll attach them. Debugging info: [New Thread 1024 (LWP 2741)] [New Thread 2049 (LWP 2742)] [New Thread 1026 (LWP 2743)] [New Thread 2051 (LWP 2744)] 0x40d7a519 in __wait4 () from /lib/i686/libc.so.6
+ Trace 24718
Assuming it's a re-enterancy / destroy bug through the control realisation process, I've committed a fix - please re-open if problems persist.
Michael: I'm still seeing the crash on the 19th sources from CVS. The reproducibility rate has now dwindled. It is less reproducible. But, I did get a crash. I also get a "Nautilus cannot be started now due to an unexpected error" dialog at times, when i keep executing the 'nautilus -q' command repeatedly. I then need to 'bonobo-slay' before starting nautilus again. Reopening bug. The new stack trace looks different from the earlier one. I'm attaching the fresh stack trace. [New Thread 1024 (LWP 9607)] [New Thread 2049 (LWP 9610)] [New Thread 1026 (LWP 9611)] [New Thread 2051 (LWP 9612)] 0x40d7f519 in __wait4 () from /lib/i686/libc.so.6
+ Trace 25308
a rather different bug, the sidebar widget not being NULLed correctly on the window. Fixed in CVS.
Crashing again :/. I'm running the 28th august sources from CVS. I repeatedly executed 'nautilus -q' and it crashed. I've set nautilus to draw my desktop. The stack trace appears different from the previous two ones. Attaching the fresh stack trace, obtained from bug buddy. Reopening bug, again :( Debugging information: [New Thread 1024 (LWP 3284)] [New Thread 2049 (LWP 3286)] [New Thread 1026 (LWP 3287)] [New Thread 2051 (LWP 3288)] 0x40c77519 in __wait4 () from /lib/i686/libc.so.6
+ Trace 26688
Michael? Is this bonobo or nautilus?
Almost certainly nautilus; it's possible that an incoming call destroys the window, before it's even finished being constructed. It's pretty hard to fix this, because of the (curious) way Nautilus does so much CORBA work in the construct time properties - in a most odd way. It looks like this needs splitting out and a ref holding over the construction process. It's almost certainly a different bug though - so it sucks to attach this trace to this old, closed bug. My attached patch might help some of the re-enterancy pain go away; it'd be good to have anyway - but I can't take the release team approval pain.
Created attachment 10785 [details] [review] a patch.
*** This bug has been marked as a duplicate of 76233 ***
Doh. My mistake.
I'm still seeing this in CVS HEAD as of 2003-03-20
I checked in the patch. Do people still see this?
*** Bug 115342 has been marked as a duplicate of this bug. ***
Closing based on lack of further feedback.