In r279093, some data structures were changed to use doubles instead of floats which resulted in a regression in the imported/w3c/web-platform-tests/html/canvas/element/path-objects/2d.path.isPointInPath.basic.html test on both GTK and WPE. My suspicion is that the actual numbers being fed into Cairo changed somehow and resulted in it returning a different result.
I am going to try to investigate this one myself.