On 02/24/2014 02:17 PM, Daniel Mack wrote: > On 02/24/2014 01:51 PM, Radoslaw Pajak wrote: >>> On 02/24/2014 12:47 PM, Daniel Mack wrote: >>> Yes, but there's one more pointer that can dangle :) >>> >>> As we're not on a fast path here, I think we can really walk the hash >>> of >>> all names in the registry of the bus, and clean up the ->activator >>> pointer. Which should also fix the bug you mentioned above. >>> >>> What about this one instead? >>> >>> >>> Thanks, >>> Daniel >> >> This does the job well also. > > Ok, thanks! > > I commited a slightly modified version now that addresses potiential > ABBA locking when unrefing the connection.
Ok, after some internal discussion, we decided to not walk all the entries in the cases you describe, as that results in O(n) behaviour. The only option to speed things up is indeed to store the name entry with the activator connection, as you proposed. Hence, I've reverted mine and commited your patch now. Thanks, Daniel _______________________________________________ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel
