You need to
before you can comment on or make changes to this bug.
Rework table cells so that the y() coordinate never lies.
Created an attachment (id=26950) [details]
Patch that not only passes but also progresses layout tests :)
(From update of attachment 26950 [details])
r=me, afaict this looks like it should help perf, but maybe i'm mistaken. anyhoo, r=me is conditional on perf being good.
also looking at the bit fields for m_intrinsicPaddingTop, m_intrinsicPaddingBottom, and m_widthChanged -- i think we should look at rearranging these to place n_widthChanged first, to reduce the liklihood that we'll end up with either horrific load alignment and/or an unnecessary word being added to RenderTableCell