Comments on Linear sRGB Commit

[Thomas E Deweese <th...@kodak.com>]

Hi all,

   This message serves as a heads up and FYI about the linear sRGB
conversion I just committed.

   First off this change affects virtually all SVG output. What does
this mean?  Well you should run regard -new on your test files after
updating, verify that there have been no regressions. 

   This verification will be non-trivial since there will be most
likely be differences between your ref image and the new output, but
you must determine if the changes are a true regression or a
correction of the existing regression (processing being done in gamma
corrected sRGB).

   When you have determined the changes are not regressions you should
regenerate your reference images.

   I have done my best to ensure that the filters all work properly
(and in the processes I have found a number of pre-existing bugs),
however with such a fundamental change I can't help but believe that
something somewhere sneaked through (especially since I can't really
trust the output of regard).

   I'm going to be working on the documentation of the new linear sRGB
functions in GraphicsUtil next week, but if you need more info on how
to best implement linear sRGB processing in a filter before I'm done
(Vincent) let me know (or look at something like ConcreteMaskRable,
simple example, or ConcreteConvolveMatrixRable, complex example).