Created attachment 188257 [details] Testcase In the attached test case, select the text. Note how random bits of shadow show up.
<rdar://problem/13071169>
Pretty sure this happens because the useClipRect flag computed here: useClipRect = !filterRenderer()->hasFilterThatMovesPixels(); isn't propagated to sublayers, so they still clip.
Created attachment 188262 [details] Patch
Comment on attachment 188262 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=188262&action=review > Source/WebCore/ChangeLog:9 > + for descendant layers that needed to not clipping to avoid artefacts A few typos in here. "needed to not clipping" and "artefacts" > LayoutTests/css3/filters/filter-repaint-shadow-layer-child.html:45 > + <!-- You should not seen part of a blue shadow inside the gray box. --> Typo: "should not seen part" => "should not see part"
Comment on attachment 188262 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=188262&action=review >> Source/WebCore/ChangeLog:9 >> + for descendant layers that needed to not clipping to avoid artefacts > > A few typos in here. "needed to not clipping" and "artefacts" Steady on, "artefacts" is the British spelling :)
Comment on attachment 188262 [details] Patch Clearing flags on attachment: 188262 Committed r143070: <http://trac.webkit.org/changeset/143070>
All reviewed patches have been landed. Closing bug.
This didn't fix it entirely.
Is this big related to bug 82813 ? Can one of the bugs be closed?
No. This is a bug with the software shadow path.
Created attachment 202999 [details] Patch
Comment on attachment 202999 [details] Patch Clearing flags on attachment: 202999 Committed r150775: <http://trac.webkit.org/changeset/150775>
*** Bug 101154 has been marked as a duplicate of this bug. ***
This bug does not seem to be fixed. I am having the exact same issue.
I meant same issue as this duplicate: https://bugs.webkit.org/show_bug.cgi?id=101154 (In reply to comment #15) > This bug does not seem to be fixed. I am having the exact same issue.
(In reply to comment #15) > This bug does not seem to be fixed. I am having the exact same issue. Do you have some reproducible steps? A test case?