GNOME Bugzilla – Bug 404474
crash starting up
Last modified: 2009-06-07 18:19:41 UTC
Distribution: Fedora Core release 6 (Zod) Gnome Release: 2.17.2 2007-01-06 (JHBuild) BugBuddy Version: 2.17.3 System: Linux 2.6.19-1.2895.fc6 #1 SMP Wed Jan 10 19:28:18 EST 2007 i686 X Vendor: The X.Org Foundation X Vendor Release: 70101000 Selinux: Permissive Accessibility: Disabled Memory status: size: 27996160 vsize: 0 resident: 27996160 share: 0 rss: 5447680 rss_rlim: 0 CPU usage: start_time: 1170638839 rtime: 0 utime: 9 stime: 0 cutime:7 cstime: 0 timeout: 2 it_real_value: 0 frequency: 0 Backtrace was generated from '/opt/gnome/bin/gnome-power-manager' Using host libthread_db library "/lib/libthread_db.so.1". [Thread debugging using libthread_db enabled] [New Thread -1209014592 (LWP 32603)] 0x00adb402 in __kernel_vsyscall ()
+ Trace 108073
Thread 1 (Thread -1209014592 (LWP 32603))
Yes, changing the g-p-m SetBrightness method to INT32 broke all the other brightness addons. I think HAL is incorrect: http://lists.freedesktop.org/archives/hal/2007-February/007156.html Richard.
So you still get this with SVN head?
Yup
Created attachment 82239 [details] [review] test patch Can you try this patch please, it might perhaps fix it.
Still crashes. Oddly reminiscent of 395174.
Created attachment 82248 [details] [review] patch Fix the return type from proxy. Also don't use the ac-adapter object before you create it. Seems to fix the crash anyway.
2007-02-10 Richard Hughes <richard@hughsie.com> * src/gpm-srv-backlight.c: (gpm_srv_backlight_init): Part of a patch from Jon McCann: Don't use the ac-adapter object before you create it. Fixes half of #404474. I've not applied the "Fix the return type from proxy" as it makes my iBook crash (generic backlight class driver) with: g_value_set_uint: assertion `G_VALUE_HOLDS_UINT (value)' failed We've also used G_TYPE_UINT as the return type for the proxy all through 2-16, without one bug about it not working, so I'm very confused. I've reviewed the HAL code for the hald_dbus method invocation and with org.freedesktop.DBus.Introspectable.Introspect and it SHOULD be a signed integer, so maybe this is a deeper bug. I'll cc davidz too.
So, I still crash at startup... we should probably have a fix for this for 2.18 right?
Well, I really don't know what to do. Changing G_TYPE_UINT to G_TYPE_INT makes g-p-m crash both on my Toshiba (i386) and by iBook (ppc), although I agree with you, it should be a int return type according to the HAL spec.
So why do your systems crash? I think you have to follow the HAL spec on this, right?
Jon, I've not had one bugreport about the proxy crash using G_TYPE_UINT and the SetBrightness generic interface for the whole of 2.16.x, and it works for me with all HAL versions. I can't imagine how many bugs we would get if I set this to G_TYPE_INT. I think we need to discuss this on the dbus or HAL list, as something fishy is going on either in dbus, dbus-glib or HAL itself, although I'm not sure what. I think we need to make a small test case for this with a small dbus server and a dbus-glib client and just test the return G_TYPE_UINT and G_TYPE_INT cases. Richard.
David has fixed the HAL bug that was triggering this. Could you install HAL from git (or use my rpms) and see if the problem is fixed? Thanks dude.
So? OBSOLETE, FIXED, INCOMPLETE?
I assume it's OBSOLETE by now. If this is still an issue, please reopen!