On Mon, 17 Nov 2025 20:32:52 GMT, Jonas Norlinder <[email protected]> 
wrote:

> Current trivial test for `TestGetTotalGcCpuTime` fails as some OSs may have a 
> slower update rate for the underlying CPU time thread counters than the code 
> anticipated, causing the test to fail which assumes that some work have been 
> performed and the count should be non-zero.
> 
> Solution: All GCs we are testing are tracing GCs. Therefore, add some marking 
> job to inflate the GC CPU time such that we can guarantee to observe a 
> non-zero value.

Thanks David and Chris for the reviews!

> I'm unclear what a GC's "slower update rate" means.

It is the OS's scheduler that needs to schedule its task to update the thread 
CPU time.

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

PR Comment: https://git.openjdk.org/jdk/pull/28362#issuecomment-3546173687

Reply via email to