On Fri, 2 Oct 2026 10:15:53 GMT, Serguei Spitsyn <[email protected]> wrote:
> Using `usleep` for 10 ms does not look completely reliable. Maybe a sleep > loop with checks would be better. But I'm not sure that it is worth it. This > approach is better than was before and testing seems to be proving it. So, > I'm okay with the fix as is. This issue is that there is nothing to check (that I'm aware of) to see if you waited long enough. If you don't wait long enough, you are at risk that some time after the PT_DETACH the process moves into the SSTOP state. That is something you can detect, and the first solution I provided did detect it and would correct it with a SIGCONT, but you still have the issue of having to wait a while to see if the process moves into the SSTOP state. It may never do so, so you have to limit how long you wait for it. ------------- PR Comment: https://git.openjdk.org/jdk/pull/33158#issuecomment-5954755141
