RFR: 8349206: j.u.l.Handler classes create deadlock risk via synchronized publish() method [v6]

[David Beaumont]

On Tue, 25 Feb 2025 01:12:55 GMT, Stuart Marks <smarks at openjdk.org> wrote:

>> David Beaumont has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   Tweaking @implNote for better rendering.
>
> src/java.logging/share/classes/java/util/logging/Handler.java line 62:
> 
>> 60:  * {@link #isLoggable(LogRecord)} while allowing code synchronized on the
>> 61:  * {@code Handler} instance to set a new log level.
>> 62:  * <p>
> 
> The above discussion is mostly internal to Handler implementations so I don't think it belongs in public javadoc. The implementation comments at line 91 below cover this (though they could be clarified a bit) but I think the mutables-volatile/unlocked-get/locked-set convention is mostly orthogonal to the deadlock issue we're trying to solve here. I'd recommend simply removing this material.
> 
> As an aside I think using volatiles this way is quite error-prone and I don't like advertising it in the javadoc.

Fair enough. Logging is a weird space in which certain synchronization/consistency strategies become quite important for performance, so I was thinking it would be useful to explain it a bit more. I agree it's not a "neat" solution, but it can can be used safely. There are definitely a few pitfalls here though.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/23491#discussion_r1969556294