PR sent - travis failed on it but the individual test runs locally - but just skips because it cant find MySQL - I have asked for some doco on how to setup up this or preferably make tests always work in a vanilla check-out without manual repository set up.
Rob On Wed, 17 Aug 2016 at 16:57 Sergio Fernández <[email protected]> wrote: > OK. So, Sebastian, could we considered MARMOTTA-652 solved or do we have > any other lateral effect to look for? > > We may add some more specific tests to MARMOTTA-651. Rob, can you send the > patch as a PR? > > Since in MARMOTTA-640 I didn't have data, the test were just pure query > evaluation; but it'd be nice to add results checking too. > > > > On Wed, Aug 17, 2016 at 12:47 AM, Rob Atkinson <[email protected]> > wrote: > > > OK - not going to get a chance to work on general windows build things > just > > yet - chasing deadlines - but if its possible to do a sanity check on the > > 651 and new issues I raised after looking at tests - i can get past the > > testing setup hurdle and hopefully look at the problem :-) > > > > If its not a high priority to fix the build, perhaps its possible to > force > > it to automatically skip certain tests on Windows in the short term? If I > > knew the best mechanism I could at least work through the build and flag > > all the tests that fail. > > > > cheers > > Rob > > > > On Wed, 17 Aug 2016 at 07:35 Sebastian Schaffert < > > [email protected]> wrote: > > > > > Infinispan works for me (on Linux). Maybe a Windows issue? Infinispan > > does > > > some low-level networking tricks. > > > > > > Sebastian Schaffert <[email protected]> schrieb am Di., > 16. > > > Aug. 2016 um 23:34 Uhr: > > > > > > > Hi, > > > > > > > > related, I managed to fix the broken tests. It seems unrelated to > > > Marmotta > > > > 640, some minor Sesame issue (const variables started with "_const" > but > > > > Marmotta expected "-const", and this broke a couple of tests. I also > > > fixed > > > > a Sesame test with invalid URIs (they are supposed to be unbound, > while > > > we > > > > threw an exception). > > > > > > > > I am wondering why Infinispan fails for you, I'll check :) > > > > > > > > Sebastian > > > > > > > > Rob Atkinson <[email protected]> schrieb am Mo., 15. Aug. 2016 > um > > > > 05:45 Uhr: > > > > > > > >> > > > >> Poked deeper into the code and worked out how to running specific > > tests > > > >> > > > >> C:\repos\github\asf\marmotta\libraries\kiwi\kiwi-sparql>mvn > > > >> -Dtest=KiWiSparqlTest#testMarmotta640_1 test > > > >> > > > >> This gets skipped because MySQL is not available - how do I force it > > to > > > >> use a particular repository in testing? > > > >> > > > >> Then I found a few worrying things: > > > >> 1) the tests do not return or check the results of any of the test > > > >> queries - only that they appear to run. > > > >> 2) There are no tests AFAICT for any CONSTRUCT queries. > > > >> > > > >> To get started I have created a stub to support CONSTRUCT queries - > it > > > >> compiles. Patch attached. The patch also updates > Marmotta640_1.sparql > > to > > > >> match the sample data. > > > >> > > > >> so to progress I need the following: > > > >> 1) doc on how to make the tests use a default backend and actually > run > > > >> 2) doc on how to use Postgres as a test backend. > > > >> 3) quick review of patch for sanity check I'm on the right track re > > > >> supporting construct > > > >> 4) ideally a worked example of how you would like a test to check > > > results > > > >> - and clean up connections afterward etc - i.e. make the existing > > tests > > > >> work the way they should. > > > >> > > > >> Cheers > > > >> Rob > > > >> > > > >> FYI > > > >> Well, its gets this far as the main build gets on Windows > > > >> > > > >> INFO] > > > >> ------------------------------------------------------------ > > ------------ > > > >> [INFO] Reactor Summary: > > > >> [INFO] > > > >> [INFO] KiWi Triplestore: Parent ........................... SUCCESS > [ > > > >> 10.579 s] > > > >> [INFO] KiWi Triplestore: Core ............................. SUCCESS > > > >> [02:52 min] > > > >> [INFO] KiWi Triplestore: Infinispan Cache Backend ......... FAILURE > > > >> [02:07 min] > > > >> [INFO] KiWi Triplestore: Hazelcast Cache Backend .......... SKIPPED > > > >> [INFO] KiWi Triplestore: EHCache Cache Backend ............ SKIPPED > > > >> [INFO] KiWi Triplestore: Versioning ....................... SKIPPED > > > >> [INFO] KiWi Triplestore: Reasoner ......................... SKIPPED > > > >> [INFO] KiWi Triplestore: SPARQL ........................... SKIPPED > > > >> [INFO] KiWi Triplestore: DB Loader ........................ SKIPPED > > > >> [INFO] > > > >> ------------------------------------------------------------ > > ------------ > > > >> [INFO] BUILD FAILURE > > > >> > > > >> The failing test is > > > >> > > > >> Tests in error: > > > >> HotRodPersistenceTest>PersistenceTest.testStoreBNode:211 ╗ JdbcSQL > > > >> Invalid val... > > > >> HotRodPersistenceTest>PersistenceTest.testStoreUriNode:134 ╗ > > JdbcSQL > > > >> Invalid v... > > > >> > > > >> > > > HotRodRepositoryConnectionTest>RepositoryConnectionTest. > > testAddStatement:261 > > > >> ╗ Repository > > > >> > > > >> > > > >> > > > >> > > > >> > > > >> > > > >> > > > >> > > > >> On Fri, 12 Aug 2016 at 05:08 Sebastian Schaffert < > > > >> [email protected]> wrote: > > > >> > > > >>> Hi Rob, > > > >>> > > > >>> Rob Atkinson <[email protected]> schrieb am Do., 11. Aug. > 2016 > > um > > > >>> 01:03 Uhr: > > > >>> > > > >>> > Hi Sebastian > > > >>> > > > > >>> > The fix isnt complete - see 651 - so it would be great if you > could > > > >>> take > > > >>> > them both into account. If you were able to point me at the > > shortcut > > > >>> to > > > >>> > run just the relevant unit tests I'd like to extend the tests to > > > >>> include > > > >>> > the 651 issue at the very least - and maybe have a look into a > fix > > if > > > >>> I can > > > >>> > isolate where the relevant code is - it wasnt obvious from just > > > >>> searching > > > >>> > across the codebase. > > > >>> > > > > >>> > (I dont really have time to do this in the context of a full > build > > > >>> which > > > >>> > takes several hours and has quite a few failing tests still on > > > >>> Windows. I > > > >>> > could run the test on a linux vm but IMHO this shouldnt be > > necessary > > > >>> either > > > >>> > and i cant really justify the setup time on that :-( ) > > > >>> > > > > >>> > > > >>> You are right, it should work perfectly fine from Windows or even > > from > > > >>> Eclipse or IntelliJ with Maven plugins. > > > >>> > > > >>> You can first to a full "mvn install -DskipTests" for the whole > > > Marmotta > > > >>> project without tests and then later change into the libraries/kiwi > > > >>> directory and run "mvn install" with tests just in that subproject. > > > This > > > >>> way you can skip all the other expensive tests. You can even just > run > > > the > > > >>> tests in the kiwi-sparql subdirectory the same way. > > > >>> > > > >>> From IntelliJ with Maven support you can just click on the Maven > tab > > on > > > >>> the > > > >>> right and you see the whole project structure. Pick the subproject, > > > right > > > >>> click on the "install" target and directly debug or run it. > > > >>> > > > >>> Hope this helps! :) > > > >>> > > > >>> Sebastian > > > >>> > > > >>> > > > >>> > > > > >>> > rob > > > >>> > > > > >>> > > > > >>> > On Thu, 11 Aug 2016 at 07:16 Sebastian Schaffert < > > > >>> > [email protected]> wrote: > > > >>> > > > > >>> > > Hi, > > > >>> > > > > > >>> > > it seems the fix to MARMOTTA-640 (SPARQL UNION) broke all UNION > > > >>> tests. At > > > >>> > > least > > > >>> > > queries 26, 28 and 29 seem to be broken. > > > >>> > > > > > >>> > > Not sure I have enough time to fix it but I'll have a look. > > > >>> > > > > > >>> > > Cheers > > > >>> > > > > > >>> > > Sebastian > > > >>> > > > > > >>> > > > > >>> > > > >> > > > > > > > > > -- > Sergio Fernández > Partner Technology Manager > Redlink GmbH > m: +43 6602747925 > e: [email protected] > w: http://redlink.co >
