On Sat Feb 21 16:30:03 2026 +0000, Jakub Woźniak wrote:
I understand the correct approach involves emitting D2D1 primitives rather than rasterizing. Given that SVG support has been absent for a long time and this implementation covers the majority of real-world UI use cases, would a bitmap-based implementation be acceptable as an interim solution while the proper implementation is developed? Or alternatively, could you point me toward which parts of the proper implementation would be most valuable to tackle first?" I would start by implementing SVG document and corresponding API. Once you have a tree, I'd imagine you would simply iterate over it, issuing appropriate context calls. The problem with interim solution like that is that it's not possible to incrementally fix it later; it introduces dependencies that we don't really need, like Cairo. I haven't looked at how much librsvg saves us.
-- https://gitlab.winehq.org/wine/wine/-/merge_requests/10156#note_130150