I think I've found a bit of a glitch with overlapping/collision detection for fine detection objects
Under certain conditions, objects are not correctly testing their collision masks against each other, if objects have been rescaled/rotated- even if these rescaling/rotations have no real impact on their collision mask. IE any scale other than 1.000, any angle other than 0. But oddly, it appears to matter where the hotspot is on the images, and will detect properly if both objects are both rescaled and rotated- just not partial combinations.
This example shows that collision detection is NOT working as intended when object "A" has a top-left hotspot and angle skewed and "B" has a center hotspot and xscale skewed
When this occurs, the collision mask of "A" apparently becomes a 1x1 pixel at its hotspot and the rest is ignored, even though the sprite is still displayed properly. It doesn't matter if its only 0.01 angle and 0.999 x-scale
yet, the collision detection works fine in the same example with either {centered hotspot} or {setting angle on the rescaled object}
it also works with boxes as expected if you disable fine collisions on either object
it also works fine if only one object has its angle or xscale set- or if both objects have both properties set
I'm just using 0.99 scale / 0.01 angle here as examples since it appears the same, I saw this bug occurring with random angles in the 360 degree range and and other scalars =/= 1.0
I don't know if I can isolate this bug any more than I already have, but its clear there's something wonky with the collision mask being calculated when scaling/angles are involved.
example file is attached
