Created attachment 178072 [details] test case This might be for Chromium only. Not tested on other platforms. scrollOffsetForFixedPosition() calculates incorrect position for fixed elements when the content (originally larger than viewport) is scaled to smaller than the viewport, causing the graphics layers of fixed position elements be placed at wrong positions. To reproduce, run the attached test case on chromium-linux. The position of the layer for the fixed element is (218, 230). Should be (10, 10).
Created attachment 178080 [details] Patch
Comment on attachment 178080 [details] Patch Attachment 178080 [details] did not pass chromium-ews (chromium-xvfb): Output: http://queues.webkit.org/results/15170670
Created attachment 178087 [details] Add missing expected file
Comment on attachment 178087 [details] Add missing expected file View in context: https://bugs.webkit.org/attachment.cgi?id=178087&action=review Added some explanations for reviewers. > Source/WebCore/page/scrolling/ScrollingCoordinator.cpp:71 > + if (maxValue <= 0) If contentsSize < visibleContentSize, return 0 directly. All other changed code should be equivalent with the original code. > Source/WebCore/page/scrolling/ScrollingCoordinator.cpp:-94 > - (contentsSize.height() - visibleContentRect.height() * frameScaleFactor) / maxOffset.height()); The above code is moved into fixedPositionScrollOffset(). Here maxOffset might be 0, and there were unnecessary floating point operations on infinity values.
@bdakin, could you please review this change? Thanks!
Comment on attachment 178087 [details] Add missing expected file R=me
Comment on attachment 178087 [details] Add missing expected file Clearing flags on attachment: 178087 Committed r137368: <http://trac.webkit.org/changeset/137368>
All reviewed patches have been landed. Closing bug.
compositing/geometry/fixed-position-composited-page-scale-smaller-than-viewport.html is failing on the WK2 bots.
(In reply to comment #9) > compositing/geometry/fixed-position-composited-page-scale-smaller-than-viewport.html is failing on the WK2 bots. Though it seems fine on my computer. Ugh.