Fix nonpremultiplied webgl toDataURL to jpeg
Created attachment 107873 [details] Patch
Comment on attachment 107873 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=107873&action=review Looks good overall. Thanks for catching this. Please update the comment when landing. r=me > LayoutTests/fast/canvas/webgl/premultiplyalpha-test.html:5 > + --> There shouldn't be any copyright headers in the WebKit layout tests; please remove this one.
Created attachment 107918 [details] Patch
Comment on attachment 107918 [details] Patch Looks good.
Comment on attachment 107918 [details] Patch Clearing flags on attachment: 107918 Committed r95482: <http://trac.webkit.org/changeset/95482>
All reviewed patches have been landed. Closing bug.
This was rolled out due to breakage of the CG port. Reopening.
Any mac-chrome breakage? I believe bug 40147 affects the CG mac ports.
Yeah, mac-chrome was broken. I'm currently working on a fix (which will hopefully fix bug 40147 as well).
Thanks John. For skia, maybe avoid the division by 255 at lines @96-98 with a loop body like ... unsigned char alpha = pixels[3]; if (alpha != 255) { *output++ = SkMulDiv255Round(pixels[0], alpha); *output++ = SkMulDiv255Round(pixels[1], alpha); *output++ = SkMulDiv255Round(pixels[2], alpha); } else { *output++ = pixels[0]; *output++ = pixels[1]; *output++ = pixels[2]; }
Created attachment 108265 [details] Patch
Comment on attachment 108265 [details] Patch Looks good to me. Nice work. Please keep an eye on the bots (since it seems the previous failure was only caught by the Chromium canaries) and once this settles down please take https://bugs.webkit.org/show_bug.cgi?id=40147 and close it as a duplicate of this one.
Comment on attachment 108265 [details] Patch Clearing flags on attachment: 108265 Committed r96000: <http://trac.webkit.org/changeset/96000>
*** Bug 40147 has been marked as a duplicate of this bug. ***