As indicated in the JBS bug, using a `PixelReader` to read a scaled image in HiDPI mode, for example an `@2x` image, to read more than one pixel will read data from the wrong location in the image, usually leading to an IOOBE.
The bug is in the `getPixelAccessor` method of Prism Image. The method correctly creates a new `Accessor` if one hasn't already been created, but then it unconditionally wraps the current `Accessor` in a `ScaledPixelAccessor` if the scale is > 1. So the second time this method is called, it created another `ScaledPixelAccessor` that wraps the first `ScaledPixelAccessor`, meaning that the indexes into the pixel array are multiplied by 4. This continues for each new call to this method. The fix is to only wrap an `Accessor` in a `ScaledPixelAccessor` the first time, when it is created. I added a test, which is run for both scaled and unscaled images, to ensure that we get the right value when reading a pixel, and that reading it a second time returns the same result. Without the fix, the new test will fail with IOOBE when the scale factor is two. Related to this, I initially commented out the wrapping in a `ScaledPixelAccessor` entirely, just to see what would happen, and none of the existing tests caught it. The new test will catch this. ------------- Commit messages: - 8258986: getColor throws IOOBE when PixelReader reads the same pixel twice Changes: https://git.openjdk.java.net/jfx/pull/389/files Webrev: https://webrevs.openjdk.java.net/?repo=jfx&pr=389&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8258986 Stats: 160 lines in 4 files changed: 157 ins; 0 del; 3 mod Patch: https://git.openjdk.java.net/jfx/pull/389.diff Fetch: git fetch https://git.openjdk.java.net/jfx pull/389/head:pull/389 PR: https://git.openjdk.java.net/jfx/pull/389
