Hi David and Felipe,
Please review my proposed fix for:
The webrev pointer, and a question, are in my recent comment there.
This makes every instance of corner radii handling I could find in
Region and NGRegion internally consistent, enforced by pre-filtering the
data through a single method in Region.
I am curious to hear from David about the interpretation I've made of
various anomalous cases of corner radii, and Felipe from general
graphics correctness of my calculations. I mention a specific question
I had while developing the fix in the Jira comment with the webrev.
I tested it with Ensemble8 and the test cases in the report. I haven't
written a specific test case that combines inset+percentage overflow
until I hear back from David (or anyone) about what the correct
interpretation is for that case...