On Tue, 17 Nov 2020 18:32:30 GMT, Jorn Vernee <jver...@openjdk.org> wrote:
> This PR sharpens the testing done by > test/jdk/java/lang/invoke/VarHandles/VarHandleTestExact.java after > @mcimadamore reported that the test was not catching an issue with memory > access var handles; namely that the implementation of withInvokeBehavior was > incorrect. > > After some debugging it turned out that the test never actually tested: > (1) going back to invoke behavior from a var handle with invoke exact behavior > (2) the memory access handle implementation of withInvoke(Exact)Behavior, due > to memory handles always being adapted. > > The patch adds testing for (1), and adds a flag to jdk.internal.foreign.Utils > to turn off the adaptation, so that we can test the 'naked' memory access var > handles as well for (2). > > I've also tried to reduce some of the code duplication by creating the higher > order doTest function, that does most of the testing (besides setting up var > handles and test values). Marked as reviewed by chegar (Reviewer). ------------- PR: https://git.openjdk.java.net/jdk/pull/1267
