|Summary:||D-Bus configuration file handling (present in .62 to .93) issues|
|Product:||dbus||Reporter:||Steev Klimaszewski <steev>|
|Component:||core||Assignee:||Havoc Pennington <hp>|
|Status:||RESOLVED NOTABUG||QA Contact:||John (J5) Palmieri <johnp>|
|i915 platform:||i915 features:|
Description Steev Klimaszewski 2006-09-19 16:38:19 UTC
If there is an empty configuration file, D-Bus will fail to start. On Gentoo, I do as such... sudo touch /etc/dbus-1/system.d/foo.conf sudo /etc/init.d/dbus restart And the output I get is Failed to start message bus: Not enough memory. This also happens with Gentoo/FreeBSD. Versions tested were .62, .91, .92, .93 Attempting to start with dbus-daemon --system --nofork also reports the same error immediately. CFLAGS used were (0.62) - -march=pentium2 -O2 -fomit-frame-pointer -pipe -fforce-addr (0.9x) - -march=pentium3 -O2 -pipe -fforce addr -march=i686 -O2 -pipe Not sure what all else info is required, simply ask and I am more than willing to provide.
Comment 1 John (J5) Palmieri 2006-09-24 10:35:20 UTC
I get Failed to start message bus: Error in file /etc/dbus-1/system.d/test.conf, line 1, column 0: no element found which is correct, though we should most likely ignore this and send a warning instead.
Comment 2 Havoc Pennington 2006-09-24 14:59:38 UTC
Unless I'm missing something I'd say NOTABUG - a config file is necessary to define a bus daemon, there is no sensible default behavior. And the minimum valid config file will have the root xml element, an empty file is not a valid xml document.
Comment 3 Steev Klimaszewski 2006-09-24 21:22:27 UTC
I would think a warning would be better than erroring. I don't know why Gentoo reports it as Not enough memory - but I have had quite a few people test it, all with the same results. It could be that checks are disabled in the ebuild, I will look into that as well tomorrow, when my internet access is a bit more reliable
Comment 4 Havoc Pennington 2006-09-25 05:31:04 UTC
An error is required because there's no way to continue; the daemon doesn't know what to do without a config file. It would not behave in any sensible way (would not even know whether to fork or not)
Comment 5 John (J5) Palmieri 2006-09-26 10:09:15 UTC
I am leaning towards not a bug other than the fact that the gentoo guys are getting out of memory errors which is not helpful in debugging the issue if it were to come up. My reasoning for keeping the abort is we can not fix user or packaging errors (would you propose we still run if someone overwrote system.conf?). A malformed policy file could indicate larger problems and would put the bus in an unkown state if the file itself should have had content in it.
Comment 6 John (J5) Palmieri 2006-10-11 08:25:11 UTC
committing as not a bug. Above states why.