This test used to fail on macOS because the core text path was used for layout and it always applied fractional metrics
On Linux the problem is that for visual bounds we use FT_Outline_Get_BBox() which operates in fixed point, whereas when obtaining the bounds from a GeneralPath (ie TextLayout.getOutline(), Java 2D uses double precision. Allowing 1/32 pixel rounding error seems like a reasonable course of action. --------- - [x] I confirm that I make this contribution in accordance with the [OpenJDK Interim AI Policy](https://openjdk.org/legal/ai). ------------- Commit messages: - 8169188 Changes: https://git.openjdk.org/jdk/pull/32931/files Webrev: https://webrevs.openjdk.org/?repo=jdk&pr=32931&range=00 Issue: https://bugs.openjdk.org/browse/JDK-8169188 Stats: 59 lines in 2 files changed: 37 ins; 1 del; 21 mod Patch: https://git.openjdk.org/jdk/pull/32931.diff Fetch: git fetch https://git.openjdk.org/jdk.git pull/32931/head:pull/32931 PR: https://git.openjdk.org/jdk/pull/32931
