RFR: 8317801: java/net/Socket/asyncClose/Race.java fails intermittently (aix)

[Alan Bateman]

On Tue, 17 Jun 2025 06:14:09 GMT, Alan Bateman <alanb at openjdk.org> wrote:

>>> I realize what is proposed here is the pending_signals patch that I attached to the JBS issue and I'm okay if we go with that. I was just curious if there is an alternative with an AIX version of UnixDispatcher.
>> 
>> Sure, I will check if we can add the AIX version of UnixDispatcher.
>
>> Sure, I will check if we can add the AIX version of UnixDispatcher.
> 
> It might be that it's not worth it but good to try again. As I said, I'm okay with the current patch, and happy that it doesn't require touching all channel implementations.

> @AlanBateman I validated the `jdk/java/net/Socket/asyncClose/Race.java` with the [preClose.patch](https://github.com/user-attachments/files/20789395/preClose.patch) and ran it for 100 iterations, all of which passed successfully. Without the fix, it is failing at 22nd iteration.

Thanks for checking. I think what we have in the current PR is good. Your additional testing with a AIX of UnixDispatcher means there is the option to refactor it out of the Unix* code but it might have to get updated periodically if there are changes.

So how about we go with what you have in the PR now. It just needs the copyright header on 2 files updated before it integrates.

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

PR Comment: https://git.openjdk.org/jdk/pull/25817#issuecomment-2986876662