[Webkit-unassigned] [Bug 42273] Convolution computation causes bad alpha channel values

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=42273


Eric Seidel <eric at webkit.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #61543|review?                     |review-
               Flag|                            |




--- Comment #3 from Eric Seidel <eric at webkit.org>  2010-07-14 14:27:59 PST ---
(From update of attachment 61543)
LayoutTests/ChangeLog:14
 +          * svg/custom/sigfpe.svg: Added.
This doesn't seem like a very useful name for a layout test.

LayoutTests/svg/custom/sigfpe.svg:5
 +  <rect x="10" y="10" width="20" height="20" fill="green"/>
historicaly 100x100 green rect has meant success in graphical tests.  I'm not sure this test even needs pixel results though.  <text> which said PASS would be sufficient.  Then you could mark it as layoutTestController.dumpAsText().

WebCore/platform/graphics/skia/SkiaUtils.cpp:89
 +          SkASSERT(false); // invalid colour channels; r,g,b should be 0 if a=0
Please use full english sentences, beginning with a capital and ending in a period.

WebCore/platform/graphics/skia/SkiaUtils.cpp:90
 +          // protect against sigfpe for production
divide by zero might be a more common name for SIGFPE.

WebCore/svg/graphics/filters/SVGFEConvolveMatrix.cpp:251
 +                  paintingData.dstPixelArray->set(pixel++, std::min(clampRGBAValue(totals[0] / m_divisor + paintingData.bias), alpha));
Bah.  This should be a function of some form.  Copy/pasting 4 lines in a row is silly.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.