On Sun, May 22, 2016 at 04:27:45PM +0300, Jonathan Afek wrote: > On Sat, May 21, 2016 at 12:39 AM, Jouni Malinen <j@xxxxx> wrote: > > I don't think this is a good idea. Verifying that broadcast delivery > > works is an important part of connectivity testing especially with all > > test cases using encrypted connection. > You are right. What about doing 10 retries for real hardware and > passing the broadcast connectivity test if one of them is received? If that means doing up to 10 retries and stopping on the first success, then yes, that's probably fine. I'd rather not add significant extra latency here for the success case, but it is fine to make that take longer if the broadcast frames do not get through and the test case would fail anyway. Please also note that there are couple of cases where hwsim_utils.test_connectivity() is used for negative testing, i.e., there is not supposed to be connectivity and the test case fails if that function does not raise an exception. Doing 10 retries there should work, but it will add some latency. That said, I think it is justifiable to do that for the non-hwsim case without more complexity since there are only couple of such negative testing cases. Though, I'd be fine with adding a parameter to the function to indicate if it is used for the negative case so that the retries could be avoided. -- Jouni Malinen PGP id EFC895FA _______________________________________________ Hostap mailing list Hostap@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/hostap
