Currently, if there's no transform on the graphics context, the CG implementation of GraphicsContext::roundToDevicePixels returns the rect passed in instead of rounding. This breaks text underlines when sub-pixel layout is enabled.
Created attachment 137601 [details] Patch
Comment on attachment 137601 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=137601&action=review Wouldn't rect.round() do the same thing? Or do we call it roundedIntRect(rect)? In either case, I would be surprised if the float -> int -> float conversion would matter, and would save adding a couple lines of code. > Source/WebCore/ChangeLog:11 > + end up being more than on3e pixel wide. This adds a static inline rounding method and uses it for on3e
Created attachment 137728 [details] Patch
Comment on attachment 137728 [details] Patch LGTM.
Comment on attachment 137728 [details] Patch Thanks Eric :)
Comment on attachment 137728 [details] Patch Clearing flags on attachment: 137728 Committed r114548: <http://trac.webkit.org/changeset/114548>
All reviewed patches have been landed. Closing bug.