On Tue, 24 Jun 2025 06:54:19 GMT, Serguei Spitsyn <sspit...@openjdk.org> wrote:
>> The test incorrectly expects that some system threads must be started in the >> `early JVMTI_PHASE_START` phase. However, there is a race here and it is >> possible that the `JVMTI_PHASE_LIVE` phase is reached before any system >> thread has been started. The fix is to wait for the first `THREAD_START` >> event at the `VM_INIT` event. >> >> The update includes some other changes and minor refactoring: >> - the file `libMAAThreadStart.c` has been renamed to `libMAAThreadStart.cpp` >> - the test library `jvmti_common.hpp` is used where it is possible >> - static variable `printdump` has been removed >> - the static block with `loadLibrary` has been removed from the >> `MAAThreadStart.java` >> >> Testing: >> - ran the modified test locally >> - TBD: submit mach5 tiers 1-5 > > Serguei Spitsyn has updated the pull request incrementally with two > additional commits since the last revision: > > - corrected the Copyright line > - small corrections and indentation fixes I don't see how waiting in VMInit can help here. Phase is set before posting VMInit events. ------------- PR Comment: https://git.openjdk.org/jdk/pull/25948#issuecomment-3001822541
