A hack was introduced in http://trac.webkit.org/changeset/28714 for an issue with offsets in an old version of cg that's long since been fixed. Since this hack is no longer necessary it's best removed.
Created attachment 105393 [details] Patch
Comment on attachment 105393 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=105393&action=review > Source/WebCore/ChangeLog:3 > + Remove offset hack from graphicscontextcg setShadow code graphicscontextcg -> GraphicsContextCG > Source/WebCore/ChangeLog:12 > + for an issue with offsets in an old version of cg that's long since been fixed. cg -> CoreGraphics.
Committed r93919: <http://trac.webkit.org/changeset/93919>
<rdar://problem/10033019>
...but conditionality put back via bug 67107.
This seems to have affected LayoutTests/fast/repaint/box-shadow-h.html
box-shadow-v.html as well.
Unless I'm misreading the tests, it appears that they're "more" correct now. The only difference is the non-blurred shadow above the last "lorem ipsum" - It now shows an offset of -5,-5 as requested, instead of having only the x and y offset component for the vertical and horizontal tests, respectively. I haven't double checked to see if those tests (using box-shadow) hit the ShadowBlur code path or not. If they do, then this patch wouldn't have affected that, correct?
Hmm, reverting my change for Lion does show those tests changing. So, again, unless I'm misreading it then this change helps us on those tests and we should rebaseline them. I can do that if others agree.
This is now broken in TOT in OS X Lion: <div style="width: 100px; height: 100px; -webkit-box-shadow: black 1px 1px; background: red;">
Reopening.
Filed bug 68041.
This issue is resolved by https://bugs.webkit.org/show_bug.cgi?id=68041. Closing. *** This bug has been marked as a duplicate of bug 68041 ***