[Odlparent-dev] New SingleFeatureTest incl. TestBundleDiag
vorburger at redhat.com
Wed Dec 14 12:11:44 UTC 2016
[Subject changed, WAS: "Re: New SingleFeatureTest : Is the controller/mdsal
wiring currently known to be broken?"]
+odlparent-dev, +Alexis (expressed support during yesterday's Kernel call):
https://git.opendaylight.org/gerrit/#/c/48901/ with the new
SingleFeatureTest incl. TestBundleDiag is now ready for wider testing /
review / merge from my side.
I have NOT verified that ALL features projects pass - because I don't know
how, any clever suggestions for how to run SingleFeatureTest on all of
autorelease, without waiting days (?) for everything to locally build and
run all tests (other than SingleFeatureTest) ?
If you want to locally test this, you can just grab c/48901 in odlparent,
and "mvn -Pq clean install" the enhanced odlparent, and cd into any
features project such as e.g. genius/features, or
netvirt/vpnservice/features to "mvn clean package" them... If you then
check target/surefire-reports/org. opendaylight.odlparent.
featuretest.SingleFeatureTest-output.txt, and grep for "TestBundleDiag",
you'll see it "proves" that all really Bundles came up (NOT just features
loaded), by asserting that they -eventually- converge to Active state. Now
break e.g. some BP wiring, just for fun, and see how it detects it...
FYI: I'm blocking a handful (5-6) Blueprint migration changes in Genius
until this is in, so it would be great if we could wrap this up, address
anything oldparent committers think is still NOK, to merge this soon-ish...
Michael Vorburger <vorburger at redhat.com> | IRC: vorburger @freenode | ~ =
On Thu, Dec 8, 2016 at 12:04 AM, Michael Vorburger <vorburger at redhat.com>
> - mailing lists
> + Robert
> On Wed, Dec 7, 2016 at 4:30 PM, Michael Vorburger <vorburger at redhat.com>
>> On Wed, Dec 7, 2016 at 12:15 PM, Michael Vorburger <vorburger at redhat.com>
>>> Background : I'm currently debugging why my new extended
>>> SingleTestFailure including checkBundleStatesDiag() still fails, now that
>>> Vratko fixed bug 6523 and Robert merged that (thanks to both of you!), e.g.
>>> on genius/features.
>>> I am finding that basically it all boils down to there not being any
>>> DOMDataBroker (and his friends) OSGi service being registered...
>>> Did something related to this area change recently and things currently
>>> just known to be completely badly broken, that badly? Unlikely. If so, is
>>> anyone already on this?
>>> Or, perhaps are the features definitions in something like
>>> genius/features incomplete for such a test to be able to work - would they
>>> need to explicitly list some implementation bundle/s which we currently do
>>> not (need to)?
>>> Or am I maybe making a conceptual mistake... this new test assumes that
>>> CSS and BP play nicely together, and wait for each other's deps, so my
>>> programmatic "diag" can "prove" that the wiring is fine - am I perhaps
>>> missing anything why this couldn't work, in an automated test? (From what
>>> I'm seeing so far, I don't think it's just a timing issue... or I'm reading
>>> the smoke signs wrong.)
>> just FYI: Actually it IS a timing issue.. getting it right when to assert
>> that all services have been resolved is not as trivial as I initially
>> implemented - but I'm getting there now.
>>> I'll continue digging, but wanted to reach out already while digging
>>> just on the chance that anyone can jump in with "that's probably
>>> because..." kind of idea.
>> I'm hoping to have a more reliable reworked version of
>> https://git.opendaylight.org/gerrit/#/c/48901/ tomorrow - and would be
>> interested in some of you testing it further then - anyone?
> FYI https://git.opendaylight.org/gerrit/#/c/48901/ is now complete from
> my side.
> Stephen, would you mind to remove your -2 tomorrow? It was mainly until
> https://bugs.opendaylight.org/show_bug.cgi?id=6523 was fixed, right? That
> is now fixed (thanks again Vratko) - and the code on c/48901
> <https://git.opendaylight.org/gerrit/#/c/48901/> is now significantly
> reworked, and more locally tested by me than the initial version. At this
> stage the -2 will confuse others and may prevent more people from locally
> testing this new much extended SingleFeatureTest?
> If nobody has a -1 on my now final propose version, then perhaps Robert or
> Stephen you could merge this soon-ish?
> Michael Vorburger <vorburger at redhat.com> | IRC: vorburger @freenode | ~ =
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Odlparent-dev