On Sat, 29 Aug 2020 00:31:51 GMT, Kevin Rushforth <k...@openjdk.org> wrote:
>> The fix and test looks good. > > I spent some time this afternoon going over the fix in more detail and doing > extensive testing on both Windows and Mac. > > I believe the fix is good. Both by inspection and by instrumenting the code, > there are no race conditions or other > problems that I can see. > The problem appears to be in the test, or else somewhere in the test harness > or the SW pipeline exposed by the test. If > I run the new CSSTest in a loop on either Mac or Windows, it will crash > intermittently. I then reverted your fix, > running the new test (which will throw an expected assertion error), and it > still crashes intermittently. My > recommendation is to use a system test, similar to what > [SVGTest.java](https://github.com/openjdk/jfx/blob/master/tests/system/src/test/java/test/javafx/scene/web/SVGTest.java) > does, rather than a unit test in the javafx.web module, which uses > `WebPage::paint`. I believe that the problem noted above is with the test harness, specifically the `WebPageShim::paint` method. I filed [JDK-8252596](https://bugs.openjdk.java.net/browse/JDK-8252596) to track fixing the tests. So the fix for _this_ PR should avoid using it, as suggested above. ------------- PR: https://git.openjdk.java.net/jfx/pull/279