[Webkit-unassigned] [Bug 54023] [chromium] Three texture-related webgl tests failing in DEBUG due to skia change

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=54023





--- Comment #6 from Mike Reed <reed at google.com>  2011-02-14 14:27:34 PST ---
In debug mode (SK_DEBUG defined for Skia), SkPackARGB32() will assert that the specified color is, in fact, legal premultiplied. i.e.

a >= r
a >= g
a >= b

I presume that this is the assert that is firing in the stack-trace.

The call site is (I think) this:


        inline void setRGBA(PixelData* dest, unsigned r, unsigned g, unsigned b, unsigned a)
        {
            if (m_premultiplyAlpha && !a)
                *dest = 0;
            else {
                if (m_premultiplyAlpha && a < 255) {
                    float alphaPercent = a / 255.0f;
                    r = static_cast<unsigned>(r * alphaPercent);
                    g = static_cast<unsigned>(g * alphaPercent);
                    b = static_cast<unsigned>(b * alphaPercent);
                }
#if PLATFORM(SKIA)
                *dest = SkPackARGB32(a, r, g, b);
#else
                *dest = (a << 24 | r << 16 | g << 8 | b);
#endif
            }
        }

The caller will apply the alpha before calling, but only if m_premultiplyAlpha is true. I don't know when this is not true (maybe it is always true), but if that is ever false, then we will skip the multiplies, which could make the assert fire.

Another cause could be that one of a,r,g,b is just out of range (doubtful), since SkPackARGB32() also asserts that each one is <= 0xFF

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