All, I have put up a pull request for OPENNLP-1285. This issue was causing OpenNLP builds to give different expected test results on OpenJDK 11 vs OpenJDK 8. (Same hardware, same code, just different JDK versions.)
The pull request changes usages of java.lang.Math to java.lang.StrictMath to give predictable floating point values across platforms. My build now passes on OpenJDK 11 (previously failing) and OpenJDK 8 (previously passing). If you have time please review the pull request at https://github.com/apache/opennlp/pull/377. Thanks, Jeff
