On Sat Sep 21 20:51:46 2024 +0000, Philipp Knechtges wrote:
TBH, I do not really know what to make of this assessment. Looking at the d2d1 tests it feels like, correctness is at best achieved approximately, and then the question arises what is the metric in which "correctness" is assessed? Unarguably, the tests as a metric are a moving target, depending on when which tests are added, thus making it a subjective choice at times. In my book, the patch at hand makes the rendering more correct in many cases. To be precise in all cases in which the individual geometries in the geometry group do not intersect, and thus the fill mode does not matter. Would you consider the patch sufficient, if it would (approximately) check for intersections and only copy if there are none, and falling back to old behavior (rendering just nothing) in case there are any (potentially false-positive) intersections?
It wouldn't consider it sufficient, it's clear that there is more to it than copying data from individual geometries. So the question is the same - what needs to be done to make it work properly, once it's understood, we could implement it partially if it's too hard to do properly.
-
Nikolay Sivov (＠nsivov)