On Mon, 23 Jan 2023 21:02:51 GMT, Chris Plummer <cjplum...@openjdk.org> wrote:
> jdb creates a ThreadStartRequest and ThreadDeathRequest so it can track all > created threads. It creates them with the SUSPEND_ALL policy. This is > unnecessary since jdb always immediately resumes all threads after doing the > thread bookkeeping. There is no interaction with the jdb user like there > would be with something like a Breakpoint event, and therefore no reason to > suspend. > > I was debugging an app that does nothing except create threads and allow them > to quickly exit. Changing the policy to SUSPEND_NONE sped up the execution of > the app by about 100x (that's 100 times faster, not 100%) This PR does not document the type of pre-integration testing that was done on these changes. We're seeing intermittent test failures in Mach5 Tier3 with 'release' bits: [JDK-8301644](https://bugs.openjdk.org/browse/JDK-8301644) com/sun/jdi/JdbStopThreadTest.java fails after JDK-8300811 @plummercj fixed JDK-8301644 a short while ago so any failures with com/sun/jdi/JdbStopThreadTest.java should subside. ------------- PR: https://git.openjdk.org/jdk/pull/12152
