GNOME Bugzilla – Bug 153673
doesn't handle dbus restarts
Last modified: 2008-06-11 14:47:51 UTC
apparently gnome-volume-manager doesn't handle dbus restarts. That's a problem for the distribution because it stops working when dbus is updated.
Agreed. Need to fix this.
Created attachment 32718 [details] [review] reconnect to dbus after dbus restart This is the patch we use in Ubuntu; it works great with hal 0.2.98 (plus some CVS patches) and g-v-m 1.0.2. However, I heard that it segfaults with hal CVS HEAD, so YMMV.
Created attachment 34721 [details] [review] new version of the patch used in the debian/ubuntu packages
Does this also segfault or does it work? If it works, we can apply it.
we use it in the debian/ubuntu packages for month and that work fine
Then lets get it in :) Robert?
That other bug, now closed, which was a showstopped was believed to be caused by this...
Oh, caused by this patch, you mean? Ah... I'll shut my yap then :)
I'm tempted to just close this...
because crashing on dbus restarting is not a bug?
what crash? there is no crash in CVS if dbus gets shut down or restarted. g-v-m shuts down cleanly when that happens... that's all it can do.
*** This bug has been marked as a duplicate of 170894 ***
I disagree with that. Some users update dbus, keep working, notice g-c-m magic doesn't happen and fill bugs. Some other restart the session and store is for a random session and from then g-v-m is not listed by the session and GNOME doesn't work as expected. How do you want them to know that updating dbus will stop g-v-m? It should handle dbus restart (the Debian/Ubuntu patch for this works fine now). If you don't agree you that you could open a dialog saying that g-v-m needs to be restarted.
then don't stop/restart dbus when it gets upgraded. you're not supposed to be doing that anyway. *** This bug has been marked as a duplicate of 170894 ***
You may agree or not on the bug but this is not a duplicate. The other bug is about a crasher due to a distribution patch, this one is a feature request.
ok, fine...
So you will keep refusing patches for this. Thanks for carring about users and other distributions around
That is totally the wrong attitude. And just wrong, anyhow, as Red Hat takes the same position as us on this: Treat DBUS upgrades as major system events. Also, this patch has been the cause of a HUGE number of bug reports. There is no crash without it. There are reports of a crash with it. It is nontrivial and imperfect to catch DBUS restarts.
Redhat does the same, and Debian/Ubuntu restart services on update on purpose ... every distribution has its reasons, that's not really a point to discuss here. We are sorry for the number of bugs but you didn't pinged any distribution guy about it nor you commented on the patches which was staying on bugzilla for month. Note that the patche has been updated on bugzilla with a fixed version. Maybe detecting dbus restarts is tricky but displaying a dialog before existing should not be a big deal. That would not make any difference for distributions which don't restart dbus and warn users on what happen for people who do.
I would like to add another perspective here. There are certain operating systems such as Solaris that have significant support for withstanding service interruptions and continuing running. As an example, the system is capable of receiving certain hardware errors, correctly restarting the software services affected by this error, and logging a detailed explanation of what happened. From this perspective it's always unacceptable that a long-running service can't be successfully restarted. And yes, this even applies to GUIs. It has nothing to do with updates. As systems get larger and the need for fault tolerance and resilience grows more and more relevant, handling restarts becomes more and more important. I understand that it might be difficult to do, but this is certainly not an invalid request.