RFR: 8349091: Charts: exception initializing in a background thread [v6]

[John Hendrikx]

On Tue, 4 Mar 2025 16:27:05 GMT, Kevin Rushforth <kcr at openjdk.org> wrote:

>> unrelated, but I would rather disallow background access to any platform properties.
>> 
>> _per the earlier email_, this might (read: will) create concurrent access when the node is not yet attached to the scene graph:
>> 
>> 
>> background thread                   fx app thread
>> 1. start creating the node
>> 2. add listener to platform prop
>> 3. keep initializing the node        <-- platform property change notification
>> 4. some more code
>> 
>> 
>> step 3 is when concurrent access occurs.
>
> There is an even more fundamental problem: JavaFX properties are not thread-safe. You cannot safely add a listener or binding to a property in one thread while another thread modifies that property's value.
> 
> So I think that deferring the binding to `Platform::accessibilityActiveProperty` is the only viable approach.There might still be an opportunity to simplify it a bit along the lines of what John mentions, as long as the code that binds to the property is guaranteed to be run on  the FX application thread.
> 
> We can be certain that setting the window property to a non-null value only happens on the FX app thread (`Stage::setScene` enforces that threading constraint).

@kevinrushforth I don't know, I would think it should be possible to create a property or property wrapper to make it thread-safe.  Just having a version with all its methods synchronized is probably sufficient.  If that saves a lot of complex logic in, what looks like, dozens of controls, then it seems like a far better solution to me.  Of course, that should only apply to special properties; I'm not suggesting making all properties synchronized -- although I think we could as, without contention (since they're supposed to be used from a single thread), it would be nearly free, and it would make all kinds of undefined behavior scenario's when used from multiple threads (by accident or otherwise) suddenly deterministic and debuggable.

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

PR Review Comment: https://git.openjdk.org/jfx/pull/1697#discussion_r1980107096