On Thu, Aug 12, 2010 at 09:44:10PM +1200, Michael Cree wrote: > On 12/08/10 07:14, Bob Tracy wrote: > >firefox-3.5.4 build with gcc/g++-4.2.4 completed moments ago, and it works. > >On to 3.5.5: not going to mess with gcc/g++-4.4.5 for the time being -- > >each build is taking somewhere in the vicinity of 6 hours on my PWS > >433au. > > Bob, what is the failure with firefox that you get? I built Firefox > 3.6.8 with the default gcc (Debian 4.4.4-8) and it is working fine on my > XP1000. > > Iceweasel 3.5.11 works fine until I select the history menu at which > point it crashes on both the XP1000 and PWS600au. > > I see bug 568211 > (http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=568211) hasn't been > addressed. That may warrant further investigation.
I'll have a look at that... Well, 3.5.5 is where the problem seems to have been introduced. Regardless of the toolchain and firefox version, the stack backtrace looks something like this (for the illegal instruction fault in libxul.so): nsComponentManagerImpl::GetService() (a few intermediate functions ...) nsNativeModuleLoader::LoadModule() (pages and pages of this same function over and over again in the trace listing) The following shows up in "dmesg" output (2.6.35 kernel): firefox-bin(11566): Instruction fault 4 pc = [<fffffffffffffff8>] ra = [<fffffc0000316538>] ps = 0000 Not tainted pc is at 0xfffffffffffffff8 ra is at do_notify_resume+0x348/0x5e0 v0 = 000000011f70ff80 t0 = 0000000000000000 t1 = fffffffffffffff2 t2 = 0000020002110820 t3 = 0000020002257808 t4 = fffffc0001051548 t5 = 6574617669725074 t6 = 6574617669720000 t7 = fffffc000d88c000 a0 = 0000000000000000 a1 = 0000000000000000 a2 = 0000000000000000 a3 = fffffffffffffff2 a4 = fffffffffffffff2 a5 = fffffffffffffff2 t8 = 000000011f70ff80 t9 = 0000020001e8ea8c t10= 00048d97d6d983c0 t11= 00000000000003e8 pv = fffffc00003178e0 at = 0000000000000001 gp = fffffffffffffff2 sp = fffffc000d88fbf8 Disabling lock debugging due to kernel taint Trace: [<fffffc0000316538>] do_notify_resume+0x348/0x5e0 [<fffffc0000310d4c>] work_pending+0x5c/0x70 [<fffffc00003394dc>] __send_signal+0x13c/0x290 [<fffffc000031085c>] entMM+0x9c/0xc0 Code: 00000000 00000000 00000000 00000000 00063301 00000432 <00001111> 000097be Repeated invocations of firefox generate the same trace, modulo a change in the last eight digits of "code", e.g., 0000149c in another instance. --Bob -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected] Archive: http://lists.debian.org/[email protected]
