https://bugzilla.novell.com/show_bug.cgi?id=342603#c1 --- Comment #1 from Felix Möller 2007-11-18 14:45:58 MST --- Created an attachment (id=183834) --> (https://bugzilla.novell.com/attachment.cgi?id=183834) gdb backtrace here is the backtrace. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c3 Wolfgang Rosenauer changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |kairo@kairo.at --- Comment #3 from Wolfgang Rosenauer 2007-11-20 12:26:42 MST --- *** Bug 343066 has been marked as a duplicate of this bug. *** https://bugzilla.novell.com/show_bug.cgi?id=343066 -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c5 Wolfgang Rosenauer changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |wolfgang@rosenauer.org, rguenther@novell.com --- Comment #5 from Wolfgang Rosenauer 2007-11-20 12:41:17 MST --- Richard, any input on that? (Or advice what to try) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c7 --- Comment #7 from Robert Kaiser 2007-11-20 13:46:46 MST --- A build without optimization options works here (the failing one I did myself was --enable-optimize="-Os -freorder-blocks -fno-reorder-functions -gstabs+ -march=pentium4 -mtune=nocona -msse2 -msse3 -mssse3 -mfpmath=sse", the failing one from Factory is probably -Os or -O2, probably also including -freorder-blocks -fno-reorder-functions -gstabs+ as we usually set those for the crash reporters to work. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c9 --- Comment #9 from Robert Kaiser 2007-11-20 19:56:53 MST --- A build compiled with the gcc43-4.3.0_20071118-4.1.i586.rpm from http://download.opensuse.org/repositories/home:/rguenther/SUSE_Factory/i586/ shows the same crash. The -gstabs+ is needed for our Mozilla crash reporters to be able to extract symbols, and AFAIK I did even get erratic info from gdb when using the default dwarf instead of gstabs+ for Mozilla, so I reverted to what we know works well with Mozilla applications - all that shouldn't matter here though, I hope. As for the line in comment #0, I wouldn't trust this to be the correct line of the crash because of "warning: Source file is more recent than executable." - and if it really would be a bug in FF, I'd wonder a lot that it only breaks with gcc 4.3 but works on build done with all other compilers tried up to now - note that this is even in Firefox 2.0 -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c11 JP Rosevear changed: What |Removed |Added ---------------------------------------------------------------------------- Severity|Normal |Blocker Status Whiteboard| |gnome-showstopper --- Comment #11 from JP Rosevear 2007-11-22 02:53:43 MST --- I can see it too on Factory. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c13 --- Comment #13 from Robert Kaiser 2007-11-22 16:05:46 MST --- Created an attachment (id=184460) --> (https://bugzilla.novell.com/attachment.cgi?id=184460) backtrace and some debugging from debug build I can confirm that recompiling only nsEventStateManager.cpp with -O0 instead of -Os (but leaving all other options as-is) makes the crash go away. The attached stacktrace and a bit of further output makes it clear that http://mxr.mozilla.org/seamonkey/source/content/events/src/nsEventStateManag... is the call where the bug is already present (mActiveContent being null, which is handed down to the state where that pointer is dereferenced). When everything is compiled as usual, this never ends up being null, but some optimization in current gcc 4.3 causes that pointer to end up as null. We never saw this with older compilers and don't see this when that file is not compiled with optimiziation, so it looks a lot like a compiler error. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c15 --- Comment #15 from Robert Kaiser 2007-11-22 17:00:57 MST --- I tracked it down by trying different sets of the flags shown in the gcc man pages as being the difference between -O1 and -O2/-Os and I found that the culprit is -ftree-vrp Without this flag, everything works fine, with this flag set on nsEventStateManager.cpp, the crash is happening. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c17 --- Comment #17 from Robert Kaiser 2007-11-26 08:51:57 MST --- Created an attachment (id=184733) --> (https://bugzilla.novell.com/attachment.cgi?id=184733) preprocessed source - nsEventStateManager.i Here's the preprocessed source you requested, but it's not for the faint of the heart: This is a 105,000-line-monster! Mozilla source easily gets large, and that file is part of the core Gecko engine, so it probably pulls in lots of stuff... And for the other question, -ftree-vrp -fwrapv really fixes the crash, while it's present with -ftree-vrp only. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c19 Michael Wolf changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |maw@novell.com --- Comment #19 from Michael Wolf 2007-11-26 12:56:28 MST --- I've submitted an updated MozillaFirefox to STABLE/Factory. Are there similar problems with other mozilla packages, such as thunderbird? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c21 --- Comment #21 from Felix Möller 2007-11-26 17:55:42 MST --- (In reply to comment #20 from Wolfgang Rosenauer)

The affected file is in Gecko and therefore in every mozilla based application. I'd expect to see the crash with all of them. I did not yet hit the problem in Thunderbird, although I am using factory's Thunderbird. So it is at least not that that prominent ...

-- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c23 Richard Guenther changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |ASSIGNED --- Comment #23 from Richard Guenther 2007-11-27 03:27:57 MST --- It looks like FindCommonAncestor() is compiled differently by the VRP pass with -fwrapv vs. -fno-wrapv. The following part is missing completely: while (offset < 0) { anc2 = anc2->GetParent(); ++offset; } -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c25 --- Comment #25 from Chema Ollés 2007-11-27 23:40:53 MST --- Hi, Since my last update from factory--> MozillaFirefox-translations-2.0.0.9-3.1 MozillaFirefox-2.0.0.9-3.1 It seems that this problems disappears Regards -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c27 --- Comment #27 from Wolfgang Rosenauer 2007-11-28 00:11:58 MST --- The packages in that repository contain the compiler flag workaround already. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c29 JP Rosevear changed: What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|bnc-team-mozilla@forge.provo.novell.com |rguenther@novell.com Severity|Blocker |Critical Status|ASSIGNED |NEW Component|Firefox |Basesystem Status Whiteboard|gnome-showstopper | --- Comment #29 from JP Rosevear 2007-11-28 18:07:28 MST --- Working here now too. Reducing to critical and re-assigning to GCC team. Thanks Richard. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=342603#c31 Richard Guenther changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |FIXED --- Comment #31 from Richard Guenther 2007-11-29 10:25:29 MST --- Fixed with Thu Nov 29 11:31:23 CET 2007 - rguenther@suse.de - Update to SVN trunk head (r130511). [#340981] - Add patch for PR33763. [#344524] - Add patch for PR34244. [#342603] - Add patch for PR34238. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.