----- Original Message -----
> On 05/28/2014 08:03 PM, Nathan Scott wrote:
> >> I think that the tests should be able to
> >> tolerate randomly discovered services on other machines. I ran into the
> >> same problem when working up a test for the new active probing. The
> >> solution ended up being more aggressive filtering. I'll have a look at
> >> these tests from the same point of view.
> > OK, sounds good. In my case here, with the Avahi tests there's remote
> > IP addresses in the output, so the qa/common.filter _host_to_ipaddr()
> > function might prove handy - I guess we could pick out the local host
> > IP that way, and discard all others? ( aaand that function looks like
> > it needs some IPv6 love :] ).
> >
> Since:
>
> - there could be an indeterminate number of remote services discovered
> - the host machine could have more than one active network interface
> - pmproxy and/or pmwebd services may or may not be running on the local
> machine at the time of the test
>
> I went the route of filtering out the urls completely and relying on the
> headers (e.g. "Discovered pmcd servers") and the exit code from pmfind
> for indication that results were found.
Cool - it was still failing for me though. :) Pls double check my changes
to 775 and 946 (deterministic output for both the case where there are and
are not remote pmproxy/pmwebd services found).
Its merged in dev now - I do see failures on the new tests (766 & 776), I
think these are ifconfig output differences between our text boxen? Can
we use pminfo/pmprobe on network.interface.{inet_addr,ipv6_addr} instead,
to get platform independence? Else we'll need updated filtering - I'll
send through my ifconfig output here privately if you prefer that route.
Thanks Dave, rest looks good to me. I'll leave it to you to followup in
subsequent commits on those other algorithmic suggestions Frank made, or
not, as you see fit.
Ohhhh ... wait, one other thought. With pmfind in "--all" mode - i.e.
probing freakin' anything and everything using every method available -
that means we keep re-probing found services, right? Should we instead
stop the probing loop, based on mechanism, as soon as *some* mechanism
finds *something*/*anything*? Otherwise we repeat ourselves(?), and as
more and more mechanisms come along, pmfind is going to get more and
more expensive...?
cheers.
--
Nathan
776.out.bad
Description: Binary data
766.out.bad
Description: Binary data
|