… versions

The hxp tests don't set up Application.current, which breaks ThreadPool. The hxp tests were always a pain to maintain anyway, so good riddance, in my opinion.

…OS anymore

… for macOS anymore"

This reverts commit d863321.

This reverts commit c93b500.

…ll other versions"

This reverts commit 76a9d3d.

…8.1.3

…nge in openfl#2658 broke certain gradients

Some gradients were not rendering correctly in Chrome or Safari (but worked in Firefox, interestingly) with OpenFL 9.4, but worked fine in OpenFL 9.3.

Keeping the new SpreadMethod implementations the same, though. They might have a similar issue, but I'm not familiar enough with how they work to test to see if it reproduces.

Relevant community thread where issue was reported: https://community.openfl.org/t/issues-after-upgrading-to-openfl-9-4-0-lime-8-2-0/14026

… intermittently and was using the default timeout of 250ms

…seif

…_scissorRect at extreme zoom levels. (openfl#2705)

* Delete Math.ceil and Math.floor for __ scissorRect.

In cases where the browser scaling ratio is very large, using Math.floor or Math.ceil will result in significant pixel errors, so it needs to be removed.

* Add uv offest.

* OpenGLRenderer: prefer ffloor() and fceil() over adding :Float type to variables

---------

Co-authored-by: Josh Tynjala <[email protected]>