On Mon, 7 Oct 2024 18:11:42 GMT, Daniel Gredler <d...@openjdk.org> wrote:
> > Perhaps you can add some extra tolerance and print out the values that
> > caused the failure
>
> Done.
>
> I noticed that your test failure was for scale 1, i.e. the smallest text size
> tested -- so I added 1 to the allowed variance (a fairly large % increase at
> this small size; text height at this scale is 6 pixels on my Linux box).
>
> > I will create a special test task that just runs your 2 tests and submit it
> > a large number of times and see if I can
> > get a better feel for it.
>
> Sounds good, let me know what you find.
The upshot was that the test failed every time on windows 64.
I will try your new tolerance.
May I suggest the following be added to the first test
diff RotatedScaledFontTest.java.orig RotatedScaledFontTest.java
45a46,52
> private static void saveImage(String name, BufferedImage image) {
> try {
> javax.imageio.ImageIO.write(image, "png", new java.io.File(name +
> ".png"));
> } catch (Exception e) {
> }
> }
>
65a73
> saveImage("bounds", image);
71a80
> saveImage("orientation", image);
77a87
> saveImage("xedge", image);
82a93
> saveImage("yedge", image);
Visualisation can help although probably it'll be an empty image in most cases
that it fails.
-------------
PR Comment: https://git.openjdk.org/jdk/pull/20993#issuecomment-2397784377
