back out my last commit.

art pointed out that timeout_set is the initializer of timeout structs.
this means that the ONQUEUE flag could be set when timeout_set is given
freshly allocated memory. my commit suddenly introduced the requirement
that you bzero a timeout before initialising it. without the bzero we
could generate false positives about the timeout being already queued.

art did produce a diff that would walk the queues when the flag was set
to see if it really was in the lists, but deraadt considers this too much
of a hit.

0 files changed, 0 insertions, 0 deletions