RFR: 8274166: Some CDS tests ignore -Dtest.cds.runtime.options

Matias Saavedra Silva matsaave at openjdk.org
Mon Apr 3 16:18:58 UTC 2023


On Sun, 2 Apr 2023 22:37:49 GMT, David Holmes <dholmes at openjdk.org> wrote:

> I'm not clear how the GC options are actually combined. Are you saying there are some tests that explicitly use a given GC, and if these also use test.cds.runtime.options, which contains a GC option, then the original value is kept and the latter is ignored? Won't that lead to redundant test execution? i.e. if we set test.cds.runtime.options to have UseZGC and then run all the CDS tests, those that explicitly use G1 will continue to use G1, but that configuration would already have been tested by a test run that didn't set test.cds.runtime.options.

Yes, that seems to be the case. There are tests that use a GC option but also take the test.cds.runtime.options=-XX+UseXYZGC which is ignored. With the original behavior, this option was fully ignored, which does raise the question of what these tests were intended to do. If they were meant to use the GC offered in the cds option, that is not the behavior of the tests, and if the tests were meant to show that the cds option would be ignored, they would indeed seem redundant. I have selected to preserve the existing behavior of the CDS option being ignored should there already be a GC selected.

-------------

PR Comment: https://git.openjdk.org/jdk/pull/13273#issuecomment-1494610909


More information about the hotspot-runtime-dev mailing list