It looks like the code in

utils-parent/utils/src/main/java/de/fau/cs/osr/utils/XmlGrammar.java

line 41

fails with Java 17 because not all unicode slashes are escaped. If I change the code from

"|[\uD800\\uDC00-\uDB7F\uDFFF]" + // #x10000-#xEFFFF

to

"|[\\uD800\\uDC00-\\uDB7F\\uDFFF]" + // #x10000-#xEFFFF

it works for me

We could check that license headers are present at build time using the license maven plugin.
This check was implemented in the Maven configuration but has not been enforced recently so I have disabled it in #4. It could be added back.

The CI should also build the project with Java 11 and 17 (LTS versions currently supported).

It would be worth documenting how the project can be built.

It would be nice to have a continuous integration service set up on this repository. That would help us assess pull requests and merge them confidently. It could potentially also deploy a snapshot of the library in the Maven Central snapshot repository, and official releases to the official Maven Central repository.

I have experience with GitHub Actions and could set that up on this repository.

Running mvn install in this repository currently fails because some files are missing a license header.

Small excerpt from the build logs:

 [INFO] Checking licenses...
Warning:  Missing header in: /home/runner/work/sweble-wikitext/sweble-wikitext/tooling/sweble-wikitext/sweble-wom3-parent/pom.xml
Warning:  Missing header in: /home/runner/work/sweble-wikitext/sweble-wikitext/tooling/sweble-wikitext/sweble-wom3-parent/sweble-wom3-swc-adapter/pom.xml

Caused by: java.lang.IllegalArgumentException: Comparison method violates its general contract!
        at java.util.ComparableTimSort.mergeLo(ComparableTimSort.java:744)
        at java.util.ComparableTimSort.mergeAt(ComparableTimSort.java:481)
        at java.util.ComparableTimSort.mergeCollapse(ComparableTimSort.java:406)
        at java.util.ComparableTimSort.sort(ComparableTimSort.java:213)
        at java.util.Arrays.sort(Arrays.java:1312)
        at de.fau.cs.osr.utils.visitor.VisitorLogic.sweepCache(VisitorLogic.java:225)
        at de.fau.cs.osr.utils.visitor.VisitorLogic.findVisit(VisitorLogic.java:205)
        at de.fau.cs.osr.utils.visitor.VisitorLogic.resolveAndVisit(VisitorLogic.java:105)

This is caused by multiple threads changing the lastUse while one thread is sorting. The synchronization on sweepCache is not enough to prevent other threads from changing lastUse. It may be enough to just attempt to partially sort the targets to identify the least frequently used, rather than calling Arrays.sort. A quick select strategy may be worth trying. (Note: this bug likely doesn't appear in older Java versions, which did not use TimSort).

A workaround for Java 8 is to do

System.setProperty("java.util.Arrays.useLegacyMergeSort", "true");

Migrated from sweble/sweble-wikitext#49