30143 2009-10-06 16:55:52 -0700 GraphicsContextSkia draws paths incorrectly 2013-04-09 12:48:36 -0700 1 1 1 Unclassified WebKit WebCore Misc. 528+ (Nightly build) All All RESOLVED WONTFIX P2 Normal --- 0 scherkus webkit-unassigned schenney silviapf oldest_to_newest 152896 0 scherkus 2009-10-06 16:55:52 -0700 I experienced this when writing RenderMediaControlsChromium. I would create a rounded rectangle path the add it via addPath() then fill it via fillPath() with alpha blending. It rendered correctly, but mousing over/using scroll wheel on the page would cause the re-fill the area, creating a different appearance. I'm thinking context isn't being saved correctly or something.. I'll try and get a better repro steps. 642987 1 silviapf 2012-06-06 17:17:15 -0700 This is happening in Skia because we only draw rounded borders when the rounding fits into the rectangle. See this code in Source/WebCore/platform/graphics/skia/GraphicsContextSkia.cpp : Function: void GraphicsContext::fillRoundedRect() if (topLeft.width() + topRight.width() > rect.width() || bottomLeft.width() + bottomRight.width() > rect.width() || topLeft.height() + bottomLeft.height() > rect.height() || topRight.height() + bottomRight.height() > rect.height()) { // Not all the radii fit, return a rect. This matches the behavior of // Path::createRoundedRectangle. Without this we attempt to draw a round // shadow for a square box. fillRect(rect, color, colorSpace); return; } Thus, when the volume slider is too close to 0% or to 100%, the rectangle gets too small to be able to draw the arcs. See bug https://bugs.webkit.org/show_bug.cgi?id=23882 . I don't quite know how to implement the rounded fillers now. Any suggestions? 642991 2 scherkus 2012-06-06 17:22:33 -0700 Oh my I filed this bug back waaaaaaay back when I initially was working on the original controls :P Are you still hitting it? 643087 3 silviapf 2012-06-06 19:18:22 -0700 I can actually get around this if I make the right side non-rounded underneath the slider thumbs. Also, I have to take the zoomFactor into account. Given all this, it might make sense to re-write this code by introducing a div for the background color and a div for the highlighted buffered range. I'll continue to try in C++ for now. :-) 644291 4 silviapf 2012-06-07 21:18:05 -0700 I think we can close this - I have a patch for the video controls that works.