[css-tables] Set table and cell widths dirty when row border width changes

third_party/WebKit/Source/core/layout/LayoutTableRow.cpp

Index: third_party/WebKit/Source/core/layout/LayoutTableRow.cpp
diff --git a/third_party/WebKit/Source/core/layout/LayoutTableRow.cpp b/third_party/WebKit/Source/core/layout/LayoutTableRow.cpp
index 1dbe48fee7c0af25b2f0403a0ae7e92f97e7f449..1a38fb609918fb173752170c05232c9a4508dd0a 100644
--- a/third_party/WebKit/Source/core/layout/LayoutTableRow.cpp
+++ b/third_party/WebKit/Source/core/layout/LayoutTableRow.cpp
@@ -81,10 +81,12 @@ void LayoutTableRow::styleDidChange(StyleDifference diff, const ComputedStyle* o
             // If the border width changes on a row, we need to make sure the cells in the row know to lay out again.
             // This only happens when borders are collapsed, since they end up affecting the border sides of the cell
             // itself.
+            table->setPreferredLogicalWidthsDirty(MarkOnlyThis);

[mstensho (USE GERRIT), 2016/05/09 08:41:53]

What about the table section? Isn't it affected by this too?

[dgrogan, 2016/05/16 23:34:17]

We don't need to call section->setPreferredLogicalWidthsDirty() because
LayoutTableSection doesn't override computePreferredLogicalWidths, it just uses
the LayoutBox implementation:
virtual void computePreferredLogicalWidths() {
clearPreferredLogicalWidthsDirty(); }

             for (LayoutBox* childBox = firstChildBox(); childBox; childBox = childBox->nextSiblingBox()) {
                 if (!childBox->isTableCell())
                     continue;
                 childBox->setChildNeedsLayout();
+                childBox->setPreferredLogicalWidthsDirty(MarkOnlyThis);

[dgrogan, 2016/05/07 00:25:54]

This works but I'm out of my element here; do you have any intuition if this
will have a significant performance impact?

[mstensho (USE GERRIT), 2016/05/09 08:41:53]

If you're worried about performance, I guess a border change on a row can only
affect the preferred widths of the first and the last cell. But I don't really
think we need to worry.

[dgrogan, 2016/05/16 23:34:17]

Ok, great.

             }
         }
     }