> OK, what I did: Awesome.
> This error looks a lot like a race condition between the setup of the test > database and the order the tests run in. Most certainly. > Are we sure: > > * the tests always run in the same order No. Resetting the DB is currently hardcoded per test, so due to subtle bugs in copy/paste of the boilerplate code it is order-dependent. Redoing this (esp. in the view of Junit 4 upgrade) is on the list. Until then we'd chase them one by one. Considering that we have hundreds of tests, this is a losing battle. > * the tests aren't running in parallel At least within one mvn execution they are sequential. And I don't think hsql/derby in-memory would allow multiple processes to connect to the same DB. Andrus On Oct 23, 2014, at 7:13 PM, Aristedes Maniatis <a...@maniatis.org> wrote: > OK, what I did: > > 1. Added Java 8 to our builds > https://builds.apache.org/job/cayenne-master/ > > 2. Set up junit test failures to be reported more clearly > https://builds.apache.org/job/Cayenne%20pull%20requests/lastBuild/testReport/ > > 3. Derby/Java 8 is now always built first and if it fails, then the other > matrix jobs are not attempted > > 4. Set up a job to build every pull request at github. That's just Java 8 > with the default db (is that derby?) > https://builds.apache.org/job/Cayenne%20pull%20requests > > > For test failures like this one: > > https://builds.apache.org/job/cayenne-master/53/cayenneTestConnection=derby,jdk=jdk-1.8.0,label=Ubuntu/testReport/org.apache.cayenne.access/DataContextClobTest/testNullClob/ > > Are we sure: > > * the tests always run in the same order > * the tests aren't running in parallel > > This error looks a lot like a race condition between the setup of the test > database and the order the tests run in. > > > Ari > > > > On 24/10/2014 2:06am, Aristedes Maniatis wrote: >> I just did that a few hours ago. Hadn't quite got around to telling everyone >> :-) >> >> I also added Java 8 to our regular builds and set up the unit tests to >> report more clearly. I was hoping to see why we get so many failures. I >> don't know if it is Jenkins or just race conditions in our tests. >> >> Ari >> >> >> On 23/10/2014 10:09pm, Andrus Adamchik wrote: >>> https://builds.apache.org/job/Cayenne%20pull%20requests/ >>> >>> Didn't know we had those setup on Jenkins. Was it us or was it done by >>> infra? >>> >>> And too bad we have ongoing Jenkins flakiness with regular builds; >>> otherwise this tool would have been very useful. >>> >>> Andrus >>> >> > > -- > --------------------------> > Aristedes Maniatis > GPG fingerprint CBFB 84B4 738D 4E87 5E5C 5EFA EF6A 7D2E 3E49 102A >
