On Mon, 06.08.12 16:47, Zbigniew Jędrzejewski-Szmek ([email protected]) wrote:
> > On 08/06/2012 04:39 PM, Lennart Poettering wrote: > > On Fri, 03.08.12 17:33, shawn ([email protected]) wrote: > > > > Applied! > > > > Note that we should be carefuly with adding additional erorr messages to > > the EXIT_SUCCESS/EXIT_FAILURE bits, since we might end up with two error > > messages instead of one, which is something we should try to avoid... > Can we specify that the policy is that the oom message is logged at the > point where the NULL return value is converted into a ENOMEM return code? It's probably not that easy. I mean, stuff such as hashmap_new() probably should never log this kind of thing on its own. So far we tried to follow the logic that "library"-like calls never log on their own, and that logging is always done only in the final tools. But this is followed only very losely. I guess one could formulate a rule like this: calls in src/shared/ should never log, but other code should, but about a thousand exceptions of this are OK... Lennart -- Lennart Poettering - Red Hat, Inc. _______________________________________________ systemd-devel mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/systemd-devel
