On Wed, Feb 10, 2021 at 11:02:49AM +0100, Ralf Habacker wrote:
The tests are designed to support an automated mode and an interactive mode.
In automated mode, only tests where the angle used is a multiple of pi/4 are performed to increase the chance of matching images created on Windows.
In interactive test mode, an image is generated for each test for visual verification.
Do you have plans to work on the implementation fixes?
It's probably better to start with that and then write specific tests that are easy to get the implementation correct for.
Adding random tests at this point will likely make the fixing the implementation harder, since more effort will have to be spent getting that particular test to pass, where in reality it may be a corner case that nobody cares about. Of course if there are apps that depend in a corner case, that's a different matter.
Huw.