Closed ctrlcctrlv closed 2 years ago
I already have a theory. I think it's a bad merger. Note how the two kinks are being caused by two times colocated corner points:
But we should rightly expect that every point on the segment which links a cap or join should be a curve point with two non-colocated handles, and that the first and last points should only be one time colocated.
It's almost certainly a merger problem.
Everything looks fine on my end pre-merger.
How are you merging @simoncozens @rosawagner?
If you don't see this before the segmentwise merger—what do you see?
Okay. Here's what we're doing administratively.
This issue is getting closed because --segmentwise
was a hack which I thought would be fine, but circular arcs are not perfectly representable as cubic Bezier curves.
So, any overlap merge would struggle with these two overlapping Bézier curves which are almost—yet not quite—equivalent.
Issue #16—not #17—is getting reopened because segmentwise is not an appropriate fix.
I will then try to fix once again the path builder method of creating stroke outlines.
Thanks for user patience.
The control points angled down (red) seem bad to me. I'm expecting a smooth top (green). Maybe unreasonably, if the underlying strokes don't have co-positioned top points.
@davelab6 they aren't. they are part of different paths.
For clarity see this exploded version:
Ah good call. I see the "4th segment of the beach ball" now.
cc @simoncozens @RosaWagner this is the correct issue for the discussion had at the tail end of the resolved #16.