Hi Goetz, Thanks for raising this issue.
I'm not surprised by MaxAdvanceIsMax test failing on some OS. The reason is that this test is very OS specific. All installed fonts are tested and a static value from each font is used for the assertion, after scale calculations. If there is a difference between what the font creator set when generating the font file and what the OS rendering library returns -considering that different FreeType library builds may return different values-, this test will fail. The internal test during development phase was more reliable because it tested a known-font only. However, shipping a font is not allowed by licenses and that's why we changed the test a bit. We should probably limit the scope somehow. Despite the test failure, I'm still confident that we did the right thing in 8218854 [1]. Kind regards, Martin.- -- [1] - https://bugs.openjdk.java.net/browse/JDK-8218854
