Table layout still needs to be done with integers to preserve rules from the spec. There were also a few places missing LayoutUnits.
Created attachment 113552 [details]
Comment on attachment 113552 [details]
View in context: https://bugs.webkit.org/attachment.cgi?id=113552&action=review
Are all those max<int> needed? If both arguments are already int, I would expect you could leave out the <int>.
Lets not leave things behind that show the churn from when we converted to LayoutUnit and back. We should only have things like <int> if they are actually needed.
> + maxWidth = max<int>(maxWidth, static_cast<int>(min(maxNonPercent, numeric_limits<LayoutUnit>::max() / 2.0f)));
> + maxWidth = max<int>(maxWidth, static_cast<int>(min(maxPercent, numeric_limits<LayoutUnit>::max() / 2.0f)));
It seems unfortunate to convert from LayoutUnit to int using static_cast. Isn’t there a better way? For example, could you just leave out the cast entirely, since the argument to max<int> is an int?
> + int cellLogicalWidth = max<int>(m_layoutStruct[pos].effectiveMinLogicalWidth, static_cast<int>(cellMinLogicalWidth * m_layoutStruct[pos].logicalWidth.value() / fixedWidth));
Same comment here about static_cast<int>. I think it can be omitted.
> + int colMaxLogicalWidth = max<int>(m_layoutStruct[pos].effectiveMaxLogicalWidth, static_cast<int>(spanMaxLogicalWidth ? cellMaxLogicalWidth * static_cast<float>(m_layoutStruct[pos].effectiveMaxLogicalWidth) / spanMaxLogicalWidth : cellMaxLogicalWidth));
Created attachment 113581 [details]
Patch for landing
Comment on attachment 113581 [details]
Patch for landing
Clearing flags on attachment: 113581
Committed r99254: <http://trac.webkit.org/changeset/99254>
All reviewed patches have been landed. Closing bug.