RFR: 8272288: Funky multiresolution image breaks graphics context

[Sergey Bylokhov]

On Thu, 19 Jan 2023 08:23:15 GMT, Jayathirth D V <jdv at openjdk.org> wrote:

>> This is the next iteration of fixing the issue of rendering "broken" image. The fix for [JDK-7183828](https://bugs.openjdk.org/browse/JDK-7183828) changed the possible IllegalArgumentException during rendering to the InvalidPipeException in assumtion that the new exception will be catched by the java2d machinery and the current-accelerated pipeline will be invalidated and software-pipeline will take care about that image.
>> 
>> But that change uncover another bug, the accelerated-pipeline invalidation by the InvalidPipeException works in D3D pipeline only, other pipelines start to use software based surfaces, but do not report about that the client code, as a result the next code does not work as expected, since contentsLost alwys return false:
>> 
>>         do {
>>             vi.validate(gc);
>>             Graphics2D g = vi.createGraphics();
>>             g.fillRect(0, 0, vi.getWidth(), vi.getHeight());
>>             g.dispose();
>>         } while (vi.contentsLost());
>> 
>> After this change the correct value will be reported by the contentsLost(), so the RepaintManager will be able to catch that.
>> But this fix uncovered another bug. That fallback code path in the RepaintManager was not tested on HiDPI screen and does not work well. I'll file a separate bug about that.
>
> test/jdk/java/awt/image/VolatileImage/ReportRenderingError.java line 61:
> 
>> 59:  *
>> 60:  * @requires (os.family == "mac")
>> 61:  * @run main/othervm -Dsun.java2d.opengl=True ReportRenderingError
> 
> I ran this test in macOS and without the fix it exits after checking only OpenGL pipeline. We dont override restoreContents() in CGLVolatileSurfaceManager/MTLVolatileSurfaceManager so its better if we can force the test to run on all pipelines on individual platforms and not exit on first failure based on jtreg instrcutions. This will make the test more robust.

This java file already contains 4 tests(each can be run/problemlists separately), one covers the default pipeline on all platforms, and the other 3 cover some platform specific just in case.

> test/jdk/java/awt/image/VolatileImage/ReportRenderingError.java line 91:
> 
>> 89:             Graphics2D g = vi.createGraphics();
>> 90:             g.setColor(Color.RED);
>> 91:             g.drawImage(image, 0, 0, null); // <- can cause InvalidPipeException
> 
> This drawImage call is expected to through InvalidPipeException all the times or is it intermittent?
> If it is intermittent why so? Please explain.
> 
> It looks like after https://github.com/openjdk/jdk/commit/8da6c8d6 we will through InvalidPipeException for all cases of custom images and not intermittently.

It is not really intermittent but depends on the java2d pipeline currently in use. InvalidPipeException is produced by the accelerated pipelines but code works fine on GDI for example.

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

PR: https://git.openjdk.org/jdk/pull/11664