Hi Bertrand Bertrand Delacretaz schrieb: > Hi, > > I was bitten by this one again, trying to run Sling tests failed due > to a stale "sling" work folder, or more precisely gave different > results when run in the launchpad/testing folder vs. top of the source > code tree. > > In my opinion, "mvn clean test" should have predictable results, > without taking previous runs into account. > > Why did we move the "sling" folders from target/sling (where mvn clean > kills them) to siblings of src folders? > Is that on purpose, or a side effect of other changes? > > I'd move them back, if people agree - having non-repeatable "mvn clean > test" cycles can be a big waste of people's time.
+1, but ... This started to surface when I redid the launchpad. One thing I did was to remove support for system properties in the Sling web app, since system properties have a lot of issues in web application scenario. I found a way to solve this problem with the maven jetty plugin, where you can inject more configuration (it requires an update of the plugin used by Sling, no big deal). I did not find a solution to this problem, yet, for integration tests, which uses the cargo plugin to launch Sling to fire the tests. Maybe we would have to start and stop Sling differently. Eg. by launching a separate process with Sling standalone and shutdown that instance after the tests. Such setup could be hooked into the pre-integration-test and post-integration-test phases. WDYT ? Regards Felix > > -Bertrand >
