On Mon, 5 Jun 2023 04:03:57 GMT, Chris Plummer <cjplum...@openjdk.org> wrote:
> The test has two issues. The first is that it assume that once the VMStart > event has arrived and one "step into" is done, it will be in the main method > of the debuggee. Once there, it determines the debuggee class name by looking > at the classtype of topmost frame. The problems is when using virtual > threads, it is actually in TestScaffold.main() at this point, so the wrong > class name is gleaned from the frame. To fix this the test just saves away > the debuggee class name, which is passed to the test as the 4th argument. > > The other issue is that the test assumes once it gets to the debuggee go() > method, there are only two frames on the stack. It's more like 16 when using > virtual threads. The test needs to account for this by counting the number of > frames when go() is entered rather than assuming it will be 2. > > Tested locally with and without the wrapper and by running tier5 svc tests. Looks good. Thanks, Serguei ------------- Marked as reviewed by sspitsyn (Reviewer). PR Review: https://git.openjdk.org/jdk/pull/14307#pullrequestreview-1465860157
