Previously the visual typesetting tools and the overlay mask used
several coordinate frames, converting between them in many places in
inconsistent ways. This elimiates all uses of coordinate frames other
than screen and script, and makes the conversion done in one place, and
only when parsing or serializing ASS.
This fixes:
- A few minor rounding errors
- Horrible brokeness when only part of the video frame is being
displayed, due to higher levels of zoom than fit onscreen or panning
the video
- Distortion of the visual typesetting tools when the combination of
overridden aspect ratio, script resolution, and video resolution did
not result in square pixels.
- Resolution-dependence of the visual typesetting tools, which resulted
in some tools becoming hard to use at zooms outside the range of
100-200%.
- Some draggable controls used the mouse's script coordinates,
resulting in noticable jerky movement at high zoom levels or when
using strange script resolutions.
Closes#966.
Originally committed to SVN as r4289.
2c876e79c3