RFR: 8274932: Render scales in EmbeddedWindow are not properly updated [v6]

[Kevin Rushforth]

On Fri, 28 Jul 2023 03:53:17 GMT, Prasanta Sadhukhan <psadhukhan at openjdk.org> wrote:

>> This is how I understand this works:
>> 
>> The `JFXPanel` makes use of an `EmbeddedWindow` that  is a subclass of `Window`.  This window listens to its `showing` property, and when it becomes visible will call `updateOutputScales`; this will in turn set the (correct?) render scales.
>> 
>> Now the above fix seems to 2nd guess this logic, and overrides these values with render scales it gets from Swing/AWT (note that it didn't do this before).
>> 
>> So, my questions:
>> 
>> - If JFXPanel never called setRenderScale before, was the JFXPanel completely broken when used on monitors that are not set at 100% scale? Did they update correctly when moved between monitors?  I get the impression that it sort of worked, except for this edge case.
>> 
>> - If it wasn't completely broken, then why is this fix needed?  Shouldn't `Window` already detect that it has become visible (with its `showing` listener) and update the render scales using the `updateOutputScales` method?  In other words, isn't this a bug that perhaps needs to be fixed in `Window`s detection of when it should be updating the output scales?
>
>> This is how I understand this works:
>> 
>> The `JFXPanel` makes use of an `EmbeddedWindow` that is a subclass of `Window`. This window listens to its `showing` property, and when it becomes visible will call `updateOutputScales`; this will in turn set the (correct?) render scales.
>> 
>> Now the above fix seems to 2nd guess this logic, and overrides these values with render scales it gets from Swing/AWT (note that it didn't do this before).
>> 
>> So, my questions:
>> 
>>     * If JFXPanel never called setRenderScale before, was the JFXPanel completely broken when used on monitors that are not set at 100% scale? Did they update correctly when moved between monitors?  I get the impression that it sort of worked, except for this edge case.
>> 
>>     * If it wasn't completely broken, then why is this fix needed?  Shouldn't `Window` already detect that it has become visible (with its `showing` listener) and update the render scales using the `updateOutputScales` method?  In other words, isn't this a bug that perhaps needs to be fixed in `Window`s detection of when it should be updating the output scales?
> 
> I think it is broken...See [JDK-8222209](https://bugs.openjdk.org/browse/JDK-8222209)
> I am not sure if it might be a bug in FX Embedded `Window` s detection logic and scale updation logic, might be FX windows-graphics/toolkit team can share their thoughts but I guess it's not as otherwise we will get lot more issues and not just in JFXPanel.
> Also, JFXPanel uses `GraphicsEnvironment.getLocalGraphicsEnvironment().                              getDefaultScreenDevice().getDefaultConfiguration().getDefaultTransform().getScaleX()` which only works for primary screen and will not work if secondary screen has different scale, so that needs to be fixed in my opinion, which is being done in this PR.

@prsadhuk @hjohn I'll take a closer look early next week, but I think the best way forward might be for Prasanta to evaluate #1189 and, if the modified fix in that PR is what we want to use, update _this_ PR to include the change that adds the call to `updateSceneState()` and removes the explicit call to `Stage::setRenderScale[XY]`. Then add John as a contributor to this PR using `/contributor add`. We can then close [JDK-8222209](https://bugs.openjdk.org/browse/JDK-8222209) as a duplicate of [JDK-8274932](https://bugs.openjdk.org/browse/JDK-8274932).

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

PR Comment: https://git.openjdk.org/jfx/pull/1171#issuecomment-1656755453