On 03/07/12 09:59, Simon McVittie wrote: > On 03/07/12 02:29, YunQiang Su wrote: >> https://bugs.freedesktop.org/show_bug.cgi?id=51657 > > I've attached a patch to that bug which will avoid the crash
The reason this appeared to regress in 1.6.2, despite the wrong assertion being much older, is that assertions are not normally enabled in a production build. In 1.6.2 they are (wrongly) enabled by default. I'll upload 1.6.2-2 to fix that. Running fcitx before dbus-launch has started the dbus-daemon will not actually *work*, though, because dbus-launch --autolaunch appears to be broken when run from Xsession.d by fcitx. This is not a regression in 1.6.2: in my testing on a wheezy/i386 virtual machine, it behaves the same in 1.6.0. This is likely to need more investigation. The problem you have in im-config is that you want to run your X session hook before dbus-launch (so that when you set environment variables there, child processes of the dbus-daemon get them), but you also want to run fcitx from your X session hook - and when you do that, it starts before the dbus-daemon, and fails to work. Unfortunately, if autolaunching did work as intended, running fcitx with D-Bus integration at that point in X session startup would still not be correct! It would start a new autolaunched dbus-daemon at that point in Xsession.d which would be used by the session, and any services activated by that dbus-daemon would not inherit environment variables set by later scripts in Xsession.d, like consolekit, gpg-agent, ssh-agent and qt-a11y. This would cause bugs similar to <https://bugs.launchpad.net/ubuntu/+source/dbus/+bug/807614> and <https://bugs.launchpad.net/unity/+bug/809900>. S -- To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
