Fix broken preferred widths optimization involving subtree layout roots

Fix broken preferred widths optimization involving subtree layout roots https://bugs.webkit.org/show_bug.cgi?id=247742 Reviewed by Alan Baradlay. Merge - https://chromium.googlesource.com/chromium/blink/+/ed902d339fbff3a62e2f6124d8d40bbad45e97c0 This patch is to remove optimization or fix introduced in 2006, which is about RenderBox::computePreferredLogicalWidths to bail when in a subtree layout. This can result in a box not updating its preferred widths even when it should in the case that its a subtree layout root that was marked for layout of its children, then subsequently marked for layout itself. * Source/WebCore/rendering/RenderBox.cpp: (RenderBox:computeLogicalWidthInFragment): Remove "subtree' layout optimization * LayoutTests/fast/layout/nested-subtree-layout-preferred-widths.html: Added Test Case * LayoutTests/fast/layout/nested-subtree-layout-preferred-widths-expected.html: Added Test Case Expectation * LayoutTests/imported/w3c/web-platform-tests/css/css-sizing/contain-intrinsic-size/auto-006-expected.txt: Updated Test Expectations * LayoutTests/imported/w3c/web-platform-tests/css/css-sizing/contain-intrinsic-size/auto-010-expected.txt: Ditto Canonical link: https://commits.webkit.org/256623@main
WebKit · Nov 14, 2022 · 06fa3e82eb2e0267284e9c57342078a070e8016c · 06fa3e8
1 parent f2e73e7
commit 06fa3e82eb2e0267284e9c57342078a070e8016c
Show file tree

Hide file tree

Showing 5 changed files with 58 additions and 20 deletions.
diff --git a/LayoutTests/fast/layout/nested-subtree-layout-preferred-widths-expected.txt b/LayoutTests/fast/layout/nested-subtree-layout-preferred-widths-expected.txt
@@ -0,0 +1,2 @@
+crbug.com/497178: This test ensures we properly recompute preferred widths for nested subtree roots. If this test is ever flaky, it should be considered failing due the the non-deterministic way we iterate over subtree layout roots.
+PASS
diff --git a/LayoutTests/fast/layout/nested-subtree-layout-preferred-widths.html b/LayoutTests/fast/layout/nested-subtree-layout-preferred-widths.html
@@ -0,0 +1,40 @@
+<!DOCTYPE html>
+<style>
+div {
+    overflow: hidden;
+}
+#root {
+    width: 200px;
+    height: 400px;
+    background: red;
+}
+#content {
+    background: green;
+    width: 100%;
+    height: 400px;
+    display: block;
+}
+#container {
+    width: 400px;
+    height: 400px;
+}
+</style>
+<script src="../../resources/check-layout.js"></script>
+<div>
+    crbug.com/497178: This test ensures we properly recompute preferred widths for
+    nested subtree roots. If this test is ever flaky, it should be considered
+    failing due the the non-deterministic way we iterate over subtree layout roots.
+</div>
+<div id="container">
+    <div id="root">
+        <div data-expected-width="400" id="content">OriginalText</div>
+    </div>
+</div>
+<script>
+document.body.offsetTop;
+var rootElement = document.getElementById("root");
+var content = document.getElementById("content");
+content.innerText = "";
+rootElement.style.width = "400px";
+checkLayout("#content");
+</script>
diff --git a/...ported/w3c/web-platform-tests/css/css-sizing/contain-intrinsic-size/auto-006-expected.txt b/...ported/w3c/web-platform-tests/css/css-sizing/contain-intrinsic-size/auto-006-expected.txt
@@ -1,19 +1,19 @@
 
 FAIL Basic usage assert_equals: Sizing normally - clientWidth expected 100 but got 1
-FAIL Last remembered size can be 0 assert_equals: Using last remembered size - clientHeight expected 0 but got 2
-FAIL Last remembered size can be set to 0 after losing display:none assert_equals: Sizing normally - clientWidth expected 100 but got 0
-FAIL Last remembered size is logical assert_equals: Sizing normally - clientWidth expected 100 but got 0
-FAIL Last remembered size survives box destruction assert_equals: Sizing normally - clientWidth expected 100 but got 0
-FAIL Last remembered size survives display type changes assert_equals: Sizing normally - clientWidth expected 100 but got 0
-FAIL Losing cis:auto removes last remembered size assert_equals: Sizing normally - clientWidth expected 100 but got 0
-FAIL Losing cis:auto removes last remembered size even if size doesn't change assert_equals: Sizing normally - clientWidth expected 100 but got 0
-FAIL Losing cis:auto removes last remembered size immediately assert_equals: Sizing normally - clientWidth expected 100 but got 0
-FAIL Losing cis:auto during display:none doesn't remove last remembered size assert_equals: Sizing normally - clientWidth expected 100 but got 0
-FAIL Lack of cis:auto during box creation removes last remembered size assert_equals: Sizing normally - clientWidth expected 100 but got 0
-FAIL Last remembered size can be removed synchronously assert_equals: Sizing normally - clientWidth expected 100 but got 0
-FAIL Disconnected element can briefly keep last remembered size assert_equals: Sizing normally - clientWidth expected 100 but got 0
-FAIL Disconnected element ends up losing last remembered size assert_equals: Sizing normally - clientWidth expected 100 but got 0
+FAIL Last remembered size can be 0 assert_equals: Using last remembered size - clientWidth expected 0 but got 1
+FAIL Last remembered size can be set to 0 after losing display:none assert_equals: Sizing normally - clientWidth expected 100 but got 1
+FAIL Last remembered size is logical assert_equals: Sizing normally - clientWidth expected 100 but got 1
+FAIL Last remembered size survives box destruction assert_equals: Sizing normally - clientWidth expected 100 but got 1
+FAIL Last remembered size survives display type changes assert_equals: Sizing normally - clientWidth expected 100 but got 1
+FAIL Losing cis:auto removes last remembered size assert_equals: Sizing normally - clientWidth expected 100 but got 1
+FAIL Losing cis:auto removes last remembered size even if size doesn't change assert_equals: Sizing normally - clientWidth expected 100 but got 1
+FAIL Losing cis:auto removes last remembered size immediately assert_equals: Sizing normally - clientWidth expected 100 but got 1
+FAIL Losing cis:auto during display:none doesn't remove last remembered size assert_equals: Sizing normally - clientWidth expected 100 but got 1
+FAIL Lack of cis:auto during box creation removes last remembered size assert_equals: Sizing normally - clientWidth expected 100 but got 1
+FAIL Last remembered size can be removed synchronously assert_equals: Sizing normally - clientWidth expected 100 but got 1
+FAIL Disconnected element can briefly keep last remembered size assert_equals: Sizing normally - clientWidth expected 100 but got 1
+FAIL Disconnected element ends up losing last remembered size assert_equals: Sizing normally - clientWidth expected 100 but got 1
 PASS Disconnected element ends up losing last remembered size even if size was 0x0
 FAIL Last remembered size survives becoming inline assert_equals: Still using previous last remembered size - clientWidth expected 100 but got 1
-FAIL Last remembered size can be set to 0x0 after losing display:inline assert_equals: Last remembered size is now 0x0 - clientHeight expected 0 but got 2
+FAIL Last remembered size can be set to 0x0 after losing display:inline assert_equals: Last remembered size is now 0x0 - clientWidth expected 0 but got 1
 
diff --git a/...ported/w3c/web-platform-tests/css/css-sizing/contain-intrinsic-size/auto-010-expected.txt b/...ported/w3c/web-platform-tests/css/css-sizing/contain-intrinsic-size/auto-010-expected.txt
@@ -1,4 +1,4 @@
 
-FAIL Last remembered size supports multiple fragments assert_equals: Using last remembered size - fragment #1 height expected 150 but got 1
-FAIL Last remembered size is updated when 2nd fragment changes size assert_equals: Using updated last remembered size - fragment #1 height expected 175 but got 1
+FAIL Last remembered size supports multiple fragments assert_equals: Using last remembered size - fragment #1 width expected 50 but got 1
+FAIL Last remembered size is updated when 2nd fragment changes size assert_equals: Using updated last remembered size - fragment #1 width expected 50 but got 1
 
diff --git a/Source/WebCore/rendering/RenderBox.cpp b/Source/WebCore/rendering/RenderBox.cpp
@@ -2607,10 +2607,6 @@ void RenderBox::computeLogicalWidthInFragment(LogicalExtentComputedValues& compu
         return;
     }
 
-    // If layout is limited to a subtree, the subtree root's logical width does not change.
-    if (element() && !view().frameView().layoutContext().isLayoutPending() && view().frameView().layoutContext().subtreeLayoutRoot() == this)
-        return;
-
     // The parent box is flexing us, so it has increased or decreased our
     // width.  Use the width from the style context.
     // FIXME: Account for block-flow in flexible boxes.