On Mon, Jul 13, 2020 at 10:06:24AM -0500, Esme Povirk (they/them) wrote:
I think it would be better not to test the drawing. I don't know how to do it reliably.
That's a bit unfortunate, but it does make things easier. The only thing I could really think of was counting the non-zero pixels in some area (regardless of their exact position) and checking that the count is greater than some threshold, which is pretty obtuse and probably worse than nothing.
I'll send another patchset with the fragile parts of the test removed shortly.