You would get flipped normals and inverted triangle winding, so be prepared to deal with that. I have added the abs() in relation to a physics related refactoring; I no longer remember exactly why but it’s possible Bullet will not play nice with negative scale. I will test if leaving it out will not cause outright crashes or asserts, if not I’ll remove it and it’ll be left at user’s responsibility to use negative scale at own risk.
EDIT: Bullet will remove objects with negative scale from simulation due to AABB violation, so scale will have to be abs():ed before passing to physics.
EDIT 2: scratch that, actually negative scale is fine, but editing the scale in editor, potentially at the point where there’s just a minus sign in the edit field, will possibly cause NaN (?) or infinite be passed as the scale value, which will cause Bullet error.