Resolved, passing in Travis now. On Fri, Aug 12, 2016 at 3:04 PM, Pat Ferrel <[email protected]> wrote:
> Alex, can you look at these unit test failures on the PR, they seem to be > in JDBCPEvents > > https://travis-ci.org/apache/incubator-predictionio/builds/151905196 > > > On Aug 12, 2016, at 1:18 PM, Pat Ferrel <[email protected]> wrote: > > Can't install unittest with pip or pip3 even though the rest of the > prerequisites work. Simply refuses to find it. Pip3 search unittest give > this: > > WebTestRunner (0.2) - Web-based interface for selectively > executing client-side Python UnitTests > unittest (0.0) - > nosetests-json-extended (0.1.0) - Create json logging output for > pythonnosetests unittest framework > > no description and 0.0 seems odd. Pip3 install unittest gives: > > Collecting unittest > Could not find a version that satisfies the requirement unittest (from > versions: ) > No matching distribution found for unittest > > Tried forcing the version to 0.0 but again no luck > > Ideas? In the meantime waiting for Travis to do it—sigh > > > On Aug 11, 2016, at 2:31 PM, Pat Ferrel <[email protected]> wrote: > > With the keystore my template-based integration test passes and I’ve put > the keystore back in. The diffs on PRs on Github seem completely wonky > right now. Git diff is trustworthy at least. > > Will try the python tests now too. > > Thanks guys, working smoothly now. > > > On Aug 11, 2016, at 11:53 AM, Donald Szeto <[email protected]> wrote: > > Hi all, > > I went ahead and pulled Pat's branch, performed a clean build, and repeated > the quick start guide of the Scala parallel recommendation template. I > could produce the same problem, and root caused it to a missing > conf/keystore.jks file. > > I think with SSL now optional, we should not be distributing a KeyStore > file. We can either quick fix it now by putting this file back, or modify > SSLConfiguration to not look for this file when SSL support is off. > > Regards, > Donald > > On Thu, Aug 11, 2016 at 8:03 AM, Chan Lee <[email protected]> wrote: > > > I want to clarify some points: > > > > 2) Deploying templates do not require SSL. If you execute > > ./make-distribution.sh using the current develop branch (provided you > > change the namespace in the template from io.prediction to > > org.apache.predictionio), you can deploy without SSL on localhost. The > > travis tests also do not use SSL and pass as can be seen in > > https://travis-ci.org/Ziemin/incubator-predictionio. > > > > 3) You can run the tests locally the same way it is run on travis > > ('python3 ${PIO_DIR}/tests/pio_tests/tests.py'). It is documented in the > > README in the same directory, but maybe this was not clear enough. If you > > have any additional questions or run into issues executing the tests, > > please let me know. > > > > Thanks, > > Chan > > > > On Thu, Aug 11, 2016 at 6:07 AM, Pat Ferrel <[email protected]> > wrote: > > > >> I’m always disinclined to commit code that has not been tested. However > >> many things in the current state of the develop branch seem broken or at > >> least too big to address in a single commit. > >> > >> 1) Style test fail in Travis that are not runnable or at least have not > >> been documented for local build. I have fixed these. > >> 2) SSL seems to still be required for deploying an engine, PIO-1 may not > >> have addressed this. Has anyone tried to build and test a template yet > >> without SSL? Unless someone can state otherwise I’m inclined to ignore > this > >> failure for now since it is out of scope for this commit. > >> 3) the test framework has been integrated into Travis but again no > >> documented way to run it locally and some of the travis errors seem > >> spurious. I am also not inclined to wait for this to be fixed unless > >> someone can point out real problems it is discovering and send > instructions > >> for how to run it locally. > >> > >> Before the extra travis tests I was able to build and train a template, > >> which fails on deploy with an SSL config error. Since the PR does not > >> strictly touch any templates I am going to have to ignore this serious > >> issue and address it in Jira. > >> > >> Unless someone vetos this I plan to reluctantly push the PR with these > >> non-trivial issues remaining. > >> > >> > >> > > > > > >
