On Tue, 10.02.15 21:26, Lennart Poettering ([email protected]) wrote: > On Tue, 10.02.15 12:06, Cristian RodrÃguez ([email protected]) wrote: > > Well, all but two of these recvmsg() invocations apply to network > sockets, which cannot receive fds, hence specifying the flag is kinda > pointless... > > For the other two cases we should probably close all fds we > receive... I mean receiving them and not doing anything with them is > not any worse or better as doing so and having O_CLOEXEC set for > them... > > That said, I think it probably makes sense to enforce the general rule > that recvmsg() must be invoked with MSG_CMSG_CLOEXEC, hence I applied > your patch and updated the CODING_STYLE to mention this.
And reverted again. It's not OK apparently to pass MSG_CMSG_CLOEXEC for to recvmsg() on raw sockets and suchlike... BTW: It is greatly appreciated to test patches before submitting them. Thanks! Lennart -- Lennart Poettering, Red Hat _______________________________________________ systemd-devel mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/systemd-devel
