On Thu, 13 Apr 2023 14:49:58 GMT, Leonid Mesnik <lmes...@openjdk.org> wrote:
> Could you please review following trivial fix which correct jvm options order > in TestScaffold. > TestScaffold combines test optionos and jtreg vm options. Before > [JDK-8304834](https://bugs.openjdk.org/browse/JDK-8304834) it passed test jvm > args as part of targetAppCommandLine. So the test VM args are added to the > jtreg vm options. But after JDK-8304834 it parse then into targetVMArgs. So > test vm args should append to jtreg vm options to override them. > > Testing: tier1-tier5 (including failing combination and testing with wrapper) > I haven't added requires to filter out the test, because e filter only > failing combination usually. So IIUC basically before [JDK-8304834](https://bugs.openjdk.org/browse/JDK-8304834) we would have: java <jtreg VM args> TestClass <test vm args> and those intended test VM args were completely ignored as the VM never saw them. Then after [JDK-8304834](https://bugs.openjdk.org/browse/JDK-8304834): java <test vm args> <jtreg VM args> TestClass and the test VM args are seen by the VM but potentially overridden by the jtreg VM args. And now we have: java <jtreg VM args> <test vm args> TestClass and things now work as planned. But it also means that since [JDK-8304834](https://bugs.openjdk.org/browse/JDK-8304834) we've started running a whole bunch of tests (how many?) in a way they have never actually run before. ------------- PR Comment: https://git.openjdk.org/jdk/pull/13462#issuecomment-1507954240