Using the VM parameter -Djdk.tracePinnedThreads has a probability of causing the application to hang

[qia nxiao]

It appears to be a similar issue. However, I have a question: why is there
a high probability of reproducing deadlock when using the VM parameter
-Djdk.tracePinnedThreads, while it is difficult to reproduce without using
this VM parameter?
This observation has also been confirmed in some of my practical
applications, where adding the parameter leads to deadlocks after a few
requests, while running without the parameter remains stable for a long
time.

Alan Bateman <Alan.Bateman at oracle.com> 于2023年8月3日周四 18:29写道：

>
>
> On 03/08/2023 09:14, qia nxiao wrote:
>
> Hello everyone, I have noticed that in certain situations, there is a
> probability that using the VM parameter -Djdk.tracePinnedThreads can cause
> the application to hang.
>
> I am not familiar with the exact reason behind this issue and I would like
> to know if there is a solution.
>
> Below is a test case that I have, and there is a certain probability of it
> getting stuck when using this parameter.
>
>
>
> https://github.com/zenghu1chen/loom-compatibility-test/blob/main/src/main/java/org/example/TracePinnedThreadsOptionTest.java
>
>
>
> This looks to the same issue that was discussed recently [1] where the
> same j.u.c. lock is being used from threads that are pinned and not pinned.
>
> -Alan
>
> [1] https://mail.openjdk.org/pipermail/loom-dev/2023-July/005989.html
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/loom-dev/attachments/20230803/1ef80071/attachment-0001.htm>