Please review the JavaFX changes needed for this issue: https://bugs.openjdk.java.net/browse/JDK-8169289
webrev: http://cr.openjdk.java.net/~ddehaven/8169289/openjfx-rt.0 For reference, the OpenJDK changes needed are here: http://cr.openjdk.java.net/~ddehaven/8169289/jdk.0/ These are being reviewed separately on core-libs-...@openjdk.java.net I've tested the following cases so far: LM_CLASS_1 : -cp some.jar main.class LM_CLASS_2 : -cp path/to/loose/classes main.class LM_JAR_1 : -jar some.jar (Main-Class) LM_JAR_2 : -jar some.jar (Main-Class + JavaFX-Application-Class) <- JAC should take precedence LM_MODULE_1 : -m module/main.class LM_MODULE_2 : -m module (main class declared in module-info) and one more in progress: LM_JAR_3 : -jar some.jar (Main-Class + JavaFX-Class-Path) <- creates new ClassLoader I've tested all combinations of having changes (or not) to openjfx and openjdk and there are no unexpected results, the only cases that fail are the LM_MODULE launch modes which fail currently. This means the changes can be pushed in any order. I will create unit tests for this, but don't want to push it since this change is split between OpenJFX and OpenJDK (else LM_MODULE modes will just fail until everything is in sync). I'll file a separate issue to push the unit tests at the appropriate time. A couple other notes: - I brought over changes for handling diacritical marks on MacOS X [1], moved class loading to a new method - I assumed not having a default preloader for the -jar case was a bug, so I fixed it -DrD- [1] https://bugs.openjdk.java.net/browse/JDK-8017248
