On Fri, Sep 13, 2013 at 1:05 PM, Ben Pfaff <b...@nicira.com> wrote: > On Thu, Sep 12, 2013 at 05:46:12PM -0700, Ethan Jackson wrote: > > Assuming we really need this ability. I don't like that we have to > > take a readlock every time we get the time. Perhaps we could could > > have a global flag which is false unless time has ever been warped. > > If it hasn't then we can simply do the xclock_gettime(). I think > > that'd add a slight race, but I can't imagine it'd matter. > > I posted a v2 that eliminates the read-lock in the common case: > http://openvswitch.org/pipermail/dev/2013-September/031879.html >
Maybe this deserves more comment. time/warp was introduced by itself initially. It is useful by itself in situations where you want to advance time without actually waiting for it to do so, but you do not care if a little extra time goes by. In fact it is probably preferable in such situations, because it allows the tests to help you discover more timing related bugs. time/stop was introduced later. It helps you write tests for things that you otherwise couldn't write tests for because they depend on very precise timing. For example, you can't really test NetFlow expiration without stopping time because a really slow machine (or one using valgrind) could go through multiple expiration periods when you expect exactly one. -- "I don't normally do acked-by's. I think it's my way of avoiding getting blamed when it all blows up." Andrew Morton
_______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
