[css-flexbox] Implementing new CSS Box Alignment values

Source/core/layout/LayoutFlexibleBox.cpp

Index: Source/core/layout/LayoutFlexibleBox.cpp
diff --git a/Source/core/layout/LayoutFlexibleBox.cpp b/Source/core/layout/LayoutFlexibleBox.cpp
index 506604ef13e73fdffb736a97f5f622e27f82ff6b..2b1454d3f6a6ad760779a134c6b0178d4933f64e 100644
--- a/Source/core/layout/LayoutFlexibleBox.cpp
+++ b/Source/core/layout/LayoutFlexibleBox.cpp
@@ -1348,6 +1348,7 @@ void LayoutFlexibleBox::alignChildren(const Vector<LineContext>& lineContexts)
             case ItemPositionCenter:
                 adjustAlignmentForChild(*child, availableAlignmentSpaceForChild(lineCrossAxisExtent, *child) / 2);
                 break;
+            case ItemPositionLastBaseline: // TODO (lajava): This feature is at-risk of being dropped during the CR period.

[cbiesinger, 2015/08/10 20:45:46]

this is also not equivalent to baseline, is it? maybe add a FIXME comment for
now.

[jfernandez, 2015/08/11 13:52:59]

Yes, you are right. It's not equivalent, indeed. I'll keep the assert and just
add the TODO comment as you have suggested.

             case ItemPositionBaseline: {
                 // FIXME: If we get here in columns, we want the use the descent, except we currently can't get the ascent/descent of orthogonal children.
                 // https://bugs.webkit.org/show_bug.cgi?id=98076
@@ -1359,15 +1360,34 @@ void LayoutFlexibleBox::alignChildren(const Vector<LineContext>& lineContexts)
                     minMarginAfterBaseline = std::min(minMarginAfterBaseline, availableAlignmentSpaceForChild(lineCrossAxisExtent, *child) - startOffset);
                 break;
             }
-            case ItemPositionLastBaseline:
-            case ItemPositionSelfStart:
-            case ItemPositionSelfEnd:
-            case ItemPositionStart:
-            case ItemPositionEnd:
+            case ItemPositionSelfStart: {
+                bool hasSameCrossStartEdges = hasOrthogonalFlow(*child) ? child->styleRef().isLeftToRightDirection() == styleRef().isLeftToRightDirection() : child->styleRef().isFlippedBlocksWritingMode() == styleRef().isFlippedBlocksWritingMode();
+                if (!hasSameCrossStartEdges)
+                    adjustAlignmentForChild(*child, availableAlignmentSpaceForChild(lineCrossAxisExtent, *child));
+                break;
+            }
+            case ItemPositionSelfEnd: {
+                bool hasSameCrossStartEdges = hasOrthogonalFlow(*child) ? child->styleRef().isLeftToRightDirection() == styleRef().isLeftToRightDirection() : child->styleRef().isFlippedBlocksWritingMode() == styleRef().isFlippedBlocksWritingMode();
+                if (hasSameCrossStartEdges)
+                    adjustAlignmentForChild(*child, availableAlignmentSpaceForChild(lineCrossAxisExtent, *child));
+                break;
+            }
             case ItemPositionLeft:
+                // Cross-axis is only paralell to container's inline-axis when column-flow is used.

[ikilpatrick, 2015/08/02 23:15:37]

s/paralell/parallel

& below.

[jfernandez, 2015/08/11 13:52:59]

Done.

+                if (isColumnFlow() && !style()->isLeftToRightDirection())
+                    adjustAlignmentForChild(*child, availableAlignmentSpaceForChild(lineCrossAxisExtent, *child));
+                break;
             case ItemPositionRight:
-                // FIXME: File a bug about implementing that. The extended grammar
-                // is not enabled by default so we shouldn't hit this codepath.
+                // Cross-axis is only paralell to container's inline-axis when column-flow is used.
+                if (isColumnFlow() && style()->isLeftToRightDirection())
+                    adjustAlignmentForChild(*child, availableAlignmentSpaceForChild(lineCrossAxisExtent, *child));
+                break;
+            case ItemPositionStart:
+                break;

[cbiesinger, 2015/08/10 20:45:46]

I don't think that's correct? start and flex-start are not necessarily the same,
right?

[jfernandez, 2015/08/11 13:52:59]

I think it's correct, but you are right when saying they are not necessarily the
same. The thing is that flex-start/end are switched based on the flex-wrap
value, which is not affecting start/end values. That logic is implemented in the
'alignmentForChild' function.

+            case ItemPositionEnd:
+                adjustAlignmentForChild(*child, availableAlignmentSpaceForChild(lineCrossAxisExtent, *child));
+                break;
+            default:
                 ASSERT_NOT_REACHED();
                 break;
             }