Am 09.02.21 um 12:55 schrieb Huw Davies:
> 
> You'd be better off testing simple cases (like rotations that
> are a multiples of pi/2, or possibly pi/4) that should be
> possible to match Windows exactly.  That way any regressions
> will get caught.
> 
done, thanks for this hint

> Dumping to a bitmap is fine for testing, but nobody is going
> to looking at them going forward.

I added dumping the image to the interactive tests.

> You should append the tests to (the admittedly long)
> draw_graphics(), they'll them get tested on a large number
> of different bitmap formats.

I think that would require a refactoring of the "simple" testing code to
match the needs of the advanced tests.

Regards
Ralf