On Wed, 25 Jun 2025 00:59:41 GMT, Serguei Spitsyn <sspit...@openjdk.org> wrote:

> > I thought if we need some sanity test for ThreadStart event with 
> > can_generate_early_vmstart capability, but looking at the spec I see no 
> > relations between them. So it's unclear what was the purpose of the test.
> 
> The test purpose was to ensure the `ThreadStart` are posted in the early 
> `VM_START` phase. The issue is that a few system threads are triggered to 
> start in the early `VM_START` but there is no guaranty they are really 
> started in that phase because of a race with the VM initializing thread.

Oh, ok. So without can_generate_early_vmstart ThreadStart events for the system 
threads in most cases are not posted at all (as they are posted in PRIMORDIAL 
phase)?
Then maybe we should keep the test, but drop the requirement that the events 
are posted in START phase? And the test needs explanation what it does and why.

-------------

PR Comment: https://git.openjdk.org/jdk/pull/25948#issuecomment-3002288544

Reply via email to