https://bugs.documentfoundation.org/show_bug.cgi?id=160036
--- Comment #9 from Patrick Luby (volunteer) <[email protected]> --- (In reply to Xisco Faulí from comment #8) > Hi Noel, > I see some new commits in > https://chromium.googlesource.com/skia/+log/refs/heads/chrome/m116 since > 2ddcf18. Would it make sense to upgrade to the latest d2c2112 to see whether > it fixes the issue ? I haven't had time to look at the code yet, but this bug looks like the XOR failures we saw when the unit tests started using Skia/Raster. IIRC, I found that the following "pixel function" strings in vcl/skia/SkiaHelper.cxx weren't doing anything even when I changed the strings: // Note that the colors are premultiplied, converting to 0-255 range // must also unpremultiply. const char* const diff = R"( vec4 main( vec4 src, vec4 dst ) { return vec4( float(int(src.r * src.a * 255.0) ^ int(dst.r * dst.a * 255.0)) / 255.0 / dst.a, float(int(src.g * src.a * 255.0) ^ int(dst.g * dst.a * 255.0)) / 255.0 / dst.a, float(int(src.b * src.a * 255.0) ^ int(dst.b * dst.a * 255.0)) / 255.0 / dst.a, dst.a ); } )"; What is interesting is the comment just below that: SkRuntimeEffect::Options opts; // Skia does not allow binary operators in the default ES2Strict mode, but that's only // because of OpenGL support. We don't use OpenGL, and it's safe for all modes that we do use. // https://groups.google.com/g/skia-discuss/c/EPLuQbg64Kc/m/2uDXFIGhAwAJ opts.maxVersionAllowed = SkSL::Version::k300; So maybe the above runtime effect is no longer supported in m116? -- You are receiving this mail because: You are the assignee for the bug.
