Hmm.
https://ci-builds.apache.org/job/Jena/job/Jena_Development_Deploy/385/
shows the same situation but that's a later commit: d5d727b
Job 384 has the "[maven-release-plugin] prepare release jena-4.7.0"
commit and that succeeded.
https://ci-builds.apache.org/job/Jena/job/Jena_Development_Test/
is OK at commit d5d727b
It looks like an initialization of tests issue for GeoSPARQL because
tests can be run in a random order.
The tests don't call Jena initialization directly. Many do touch Jena
code in setupClass but if there is already class initialization running,
recursive initialization is disabled (by the JDK).
Running the first test class in the pastebin - AngleFFTest - on it's own
fails on the ARQ constant for the PropertyFunctionRegistry
That isn't NoClassDefFoundError but broken class initialization may
leave things in a inconsistent state.
GeoSPARQL does have a serviceloader init that looks right.
I think we need to fix this and do RC2.
Andy
PS
It is only tests.
System initialization is being called. So geosparql must be partially
initializing, triggering JenaSystem.init but during it's own class init.
I love it when "static final Object abc = new X()" is null.
On 24/12/2022 17:16, Bruno Kinoshita wrote:
Hi Andy,
I just came back home and booted up Ubuntu, opened the browser and command
line and tried these commands. But I did use Eclipse in the past, so tried
these now:
find . -name work -type d -exec rm -rf {} \;
find . -name target -type d -exec rm -rf {} \;
mvn clean
mvn install
Same error.
mvn install -rf :jena-geosparql
also gives the same errors.
Bruno
