On 3/9/20 6:29 am, Martin Desruisseaux wrote:
Thanks for the analysis. I can volunteer for creating the CSR if I can have
some pointers (e.g. template) to follow?
Let's start with the fix/test. Send me a patch and I will upload it to the
cr.openjdk, so will be able to send an official review request.
Indeed, I though that Graphics2D may have "if (image instanceof BufferedImage)"
optimizations that hides the issue. My plan was to create a test with a RenderedImage
wrapping a BufferedImage, which would prevent Graphics2D to use such fast track, and see
how it behaves. What I can said however is that Graphics2D works well with multi-tiled
images compliant with the specification, such as javax.media.jai.TiledImage [1].
For the formula used for "pixel coordinates to tile indices" calculation, we
can use source code of [2] as a reference.
[1]https://docs.oracle.com/cd/E17802_01/products/products/java-media/jai/forDevelopers/jai-apidocs/javax/media/jai/TiledImage.html
[2]https://docs.oracle.com/cd/E17802_01/products/products/java-media/jai/forDevelopers/jai-apidocs/javax/media/jai/PlanarImage.html#XToTileX(int,%20int,%20int)
Regards,
Martin
Le 09/03/2020 à 01:29, Sergey Bylokhov a écrit :
I have checked the history of these methods and their usage.
These methods were added to the BufferedImage back in 1997 when the
generic support of rendered images was added. Since then these methods
were used in the implementation of Graphics2D.drawRenderedImage(), but
in fact for BufferedImage they are never called [1]. I think it is
possible to change getTileGridXOffset()/getTileGridYOffset to return 0.
But it will be necessary to create a CSR, since the change has
compatibility impact.
[1]
https://github.com/openjdk/client/blob/master/src/java.desktop/share/classes/sun/java2d/SunGraphics2D.java#L2662
--
Best regards, Sergey.
