On Sun, 6 Sep 2026 08:32:16 GMT, Laurent Bourgès <[email protected]> wrote:
>> Although I think I actually fixed the original concern already with a >> previous commit (accounting for the stroke), I've refactored this to be >> simpler now. >> >> I now use the actual rendered pixels (via `SWGraphics`) to determine the >> dirty rect. A deeper investigation turns out that the system basically >> already knows this, and it is more accurate than the estimation code that I >> had before (ie. expand by stroke/miter/etc). It only required relatively >> minor changes, and removes a lot of "guess" code from `SWDrawingContext`. >> >> I've confirmed this also by supporting italic fonts in the font stubs (a >> simple guess there won't take overhang of glyphs into account). The test >> failed before and now passes. > > Yes I think marlin renderer determined the appropriate shape's bounds in > pixels and subpixels... > See > https://github.com/openjdk/jfx/blob/6670f897b0614dfc6ae9f2b08162a56006870870/modules/javafx.graphics/src/main/java/com/sun/marlin/Renderer.java#L1426 > > It is trivial to share the info through jfx api... See https://github.com/openjdk/jfx/blob/6670f897b0614dfc6ae9f2b08162a56006870870/modules/javafx.graphics/src/main/java/com/sun/prism/impl/shape/DMarlinRasterizer.java#L92 ------------- PR Review Comment: https://git.openjdk.org/jfx/pull/1969#discussion_r3943475625
