RFR: 8190329: [macos] Swing InterOp Platform.exit() crash [v2]

[Kevin Rushforth]

On Wed, 28 Aug 2024 18:22:03 GMT, Sergey Bylokhov <serb at openjdk.org> wrote:

>> Prasanta Sadhukhan has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   Reattach thread if AWT is the owner of NSApplication
>
> src/java.desktop/macosx/native/libosxapp/ThreadUtilities.m line 67:
> 
>> 65: }
>> 66: 
>> 67: + (JNIEnv*)getJNIEnv {
> 
> The cached version of 'getJNIEnv' was added for performance reasons to skip `attachCurrentThread` if we know for sure that initialization has already completed. But is it still necessary to track all this flags now, maybe the `attachCurrentThread` now smart enough to check that the current thread was already attached, or it is just fast enough?

It seems best to make minimal changes in the normal case where AWT creates the toolkit, which this PR does.

You raise a good question as to whether we could get rid of the caching entirely. That seems like something to consider in a follow-up. We would need to do measure the performance impact.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/20688#discussion_r1736452484