Timeline

Aug 25, 2019:

11:16 PM Changeset in webkit [249093] by Simon Fraser

• 19 edits
  1 add in trunk/Source/WebKit

[iOS WK2] Make a strongly-typed TransactionID to replace uint64_t transactionIDs
​https://bugs.webkit.org/show_bug.cgi?id=199983

Reviewed by Dean Jackson.

Add TransactionID which is a MonotonicObjectIdentifier<TransactionIDType>. This is modeled
after ObjectIdentifier<>, but we can't use that because it doesn't have a guarantee of
values always increasing by 1 (all derived classes share the same value source). Also, we
need a per-RemoteLayerTreeDrawingArea set of values, but a static seed would cause values to
be incremented by all RemoteLayerTreeDrawingAreas in a WebProcess.

Replace all the bare uint64_t with TransactionID, fixing message generation codegen.

• Scripts/webkit/messages.py:
• Shared/RemoteLayerTree/RemoteLayerTreeTransaction.h:

(WebKit::RemoteLayerTreeTransaction::transactionID const):
(WebKit::RemoteLayerTreeTransaction::setTransactionID):

• Shared/TransactionID.h: Added.

(WebKit::MonotonicObjectIdentifier::MonotonicObjectIdentifier):
(WebKit::MonotonicObjectIdentifier::isHashTableDeletedValue const):
(WebKit::MonotonicObjectIdentifier::encode const):
(WebKit::MonotonicObjectIdentifier::decode):
(WebKit::MonotonicObjectIdentifier::operator== const):
(WebKit::MonotonicObjectIdentifier::operator> const):
(WebKit::MonotonicObjectIdentifier::operator>= const):
(WebKit::MonotonicObjectIdentifier::operator< const):
(WebKit::MonotonicObjectIdentifier::operator<= const):
(WebKit::MonotonicObjectIdentifier::operator!= const):
(WebKit::MonotonicObjectIdentifier::increment):
(WebKit::MonotonicObjectIdentifier::next const):
(WebKit::MonotonicObjectIdentifier::toUInt64 const):
(WebKit::MonotonicObjectIdentifier::operator bool const):
(WebKit::MonotonicObjectIdentifier::loggingString const):
(WebKit::MonotonicObjectIdentifier::hashTableDeletedValue):
(WebKit::MonotonicObjectIdentifier::isValidIdentifier):
(WebKit::operator<<):

• Shared/VisibleContentRectUpdateInfo.h:

(WebKit::VisibleContentRectUpdateInfo::VisibleContentRectUpdateInfo):
(WebKit::VisibleContentRectUpdateInfo::lastLayerTreeTransactionID const):

• UIProcess/API/Cocoa/WKWebView.mm:

(-[WKWebView _processWillSwapOrDidExit]):

• UIProcess/RemoteLayerTree/RemoteLayerTreeDrawingAreaProxy.h:
• UIProcess/RemoteLayerTree/RemoteLayerTreeDrawingAreaProxy.messages.in:
• UIProcess/RemoteLayerTree/RemoteLayerTreeDrawingAreaProxy.mm:

(WebKit::RemoteLayerTreeDrawingAreaProxy::willCommitLayerTree):
(WebKit::RemoteLayerTreeDrawingAreaProxy::commitLayerTree):

• UIProcess/WebPageProxy.cpp:

(WebKit::WebPageProxy::resetState):

• UIProcess/WebPageProxy.h:
• UIProcess/ios/WKContentViewInteraction.h:
• UIProcess/ios/WKContentViewInteraction.mm:

(-[WKContentView cleanupInteraction]):

• UIProcess/ios/WebPageProxyIOS.mm:

(WebKit::WebPageProxy::commitPotentialTap):
(WebKit::WebPageProxy::handleTap):

• WebKit.xcodeproj/project.pbxproj:
• WebProcess/WebPage/RemoteLayerTree/RemoteLayerTreeDrawingArea.h:

(WebKit::RemoteLayerTreeDrawingArea::nextTransactionID const):
(WebKit::RemoteLayerTreeDrawingArea::lastCommittedTransactionID const):
(WebKit::RemoteLayerTreeDrawingArea::takeNextTransactionID):

• WebProcess/WebPage/WebFrame.h:

(WebKit::WebFrame::firstLayerTreeTransactionIDAfterDidCommitLoad const):
(WebKit::WebFrame::setFirstLayerTreeTransactionIDAfterDidCommitLoad):

• WebProcess/WebPage/WebPage.h:
• WebProcess/WebPage/WebPage.messages.in:
• WebProcess/WebPage/ios/WebPageIOS.mm:

(WebKit::WebPage::handleTap):
(WebKit::WebPage::handlePotentialDoubleTapForDoubleClickAtPoint):
(WebKit::WebPage::commitPotentialTap):

6:51 PM Changeset in webkit [249092] by Fujii Hironori

• 2 edits in trunk/Source/WTF

Regression(r248533) Assertion hit in isMainThread() for some clients using WTF because the main thread is not initialized
​https://bugs.webkit.org/show_bug.cgi?id=201083
<rdar://problem/54651993>

Unreviewed build fox for Windows.

• wtf/win/MainThreadWin.cpp:

(WTF::isMainThreadInitialized): Added.

Aug 24, 2019:

7:14 PM Changeset in webkit [249091] by Simon Fraser

• 3 edits in trunk/Source/WebCore

Have RenderLayer::calculateClipRects() use offsetFromAncestor() when possible
​https://bugs.webkit.org/show_bug.cgi?id=201066

Reviewed by Dean Jackson.

offsetFromAncestor() is a layer tree walk, so faster than localToContainerPoint(),
but we can't use it when there are transforms on the layer and intermediates up
to the ancestor.

canUseConvertToLayerCoords() was trying to answer the question about whether it's
OK to use offsetFromAncestor() (which calls convertToLayerCoords() internally), but
it has insufficient information to make a determination. Leave this issue alone, but
at least rename canUseConvertToLayerCoords().

• rendering/RenderLayer.cpp:

(WebCore::RenderLayer::updateLayerPositions):
(WebCore::RenderLayer::calculateClipRects const):

• rendering/RenderLayer.h:

2:42 PM Changeset in webkit [249090] by Simon Fraser

• 3 edits
  2 adds in trunk

Page crashes under CGPathAddUnevenCornersRoundedRect
​https://bugs.webkit.org/show_bug.cgi?id=201117

Reviewed by Dean Jackson.
Source/WebCore:

Fix crash on ​https://onehtmlpagechallenge.com/entries/pure-css-still-life-water-lemon.html
We were passing CG radius values where the sum of two radii was greater than the height or
width, caused by rounding when converting from floats to doubles.

Test: fast/borders/renderable-uneven-rounded-rects.html

• platform/graphics/cg/PathCG.cpp:

(WebCore::Path::platformAddPathForRoundedRect):

LayoutTests:

• fast/borders/renderable-uneven-rounded-rects-expected.txt: Added.
• fast/borders/renderable-uneven-rounded-rects.html: Added.

11:21 AM Changeset in webkit [249089] by Devin Rousso

• 5 edits in trunk

Web Inspector: "Copy Rule" menu item does not propagate comments properly
​https://bugs.webkit.org/show_bug.cgi?id=201095

Reviewed by Joseph Pecoraro.

Source/WebInspectorUI:

• UserInterface/Models/CSSProperty.js:

(WI.CSSProperty.prototype.commentOut):
(WI.CSSProperty.prototype.get formattedText):
Wrap the text in /* ${text} */ if the WI.CSSProperty isn't enabled (e.g. commented out).

LayoutTests:

• inspector/css/generateCSSRuleString.html:
• inspector/css/generateCSSRuleString-expected.txt:

10:35 AM Changeset in webkit [249088] by Simon Fraser

• 3 edits in trunk/Source/WebCore

RenderLayer::updateLayerPositions() doesn't propagate the ancestor flags correctly
​https://bugs.webkit.org/show_bug.cgi?id=201115

Reviewed by Zalan Bujtas.

When an updateLayerPositions() traversal starts at a non-root layer, we failed to populate
the ancestor-related UpdateLayerPositionsFlag flags, leaving layers with missing flags
(e.g. the m_hasTransformedAncestor flag). This is detected by the patch in bug 201066.

Fix by having updateLayerPositionsAfterStyleChange() and updateLayerPositionsAfterLayout()
initialize the flags from the parent layer.

This is a behavior change not detected by any existing test, but will be exercised once
the patch from bug 201066 lands.

• rendering/RenderLayer.cpp:

(WebCore::RenderLayer::flagsForUpdateLayerPositions):
(WebCore::RenderLayer::updateLayerPositionsAfterStyleChange):
(WebCore::RenderLayer::updateLayerPositionsAfterLayout):
(WebCore::outputPaintOrderTreeLegend):
(WebCore::outputPaintOrderTreeRecursive): Log hasTransformedAncestor().

• rendering/RenderLayer.h:

9:02 AM Changeset in webkit [249087] by Chris Dumez

• 2 edits in trunk/Source/WebKit

Make CacheStorageEngineCaches's decodeCachesNames() more robust against bad input data
​https://bugs.webkit.org/show_bug.cgi?id=201102

Reviewed by Antti Koivisto.

Use Vector::tryReserveCapacity() instead of Vector::reserveInitialCapacity() in CacheStorage::decodeCachesNames()
since the size is read from disk and thus cannot be trusted. If the size is too large, reserveInitialCapacity()
would end up crashing the network process. Now, we merely discard the data if tryReserveCapacity() fails because
the size is too large.

• NetworkProcess/cache/CacheStorageEngineCaches.cpp:

(WebKit::CacheStorage::decodeCachesNames):

6:36 AM Changeset in webkit [249086] by Alan Bujtas

• 2 edits in trunk/Source/WebCore

[LFC][TFC] Add section renderers to the layout tree (THEAD/TBODY/TFOOT)
​https://bugs.webkit.org/show_bug.cgi?id=201114
<rdar://problem/54664992>

Reviewed by Antti Koivisto.

Section renderers (THEAD/TBODY/TFOOT) are direct children of the RenderTable. Let's include them in the layout tree as well.

• layout/layouttree/LayoutTreeBuilder.cpp:

(WebCore::Layout::TreeBuilder::createTableStructure):
(WebCore::Layout::outputInlineRuns):

6:34 AM Changeset in webkit [249085] by Alan Bujtas

• 4 edits in trunk/Source/WebCore

[LFC] Box::isAnonymous() can not rely on the lack of ElementType.
​https://bugs.webkit.org/show_bug.cgi?id=201106
<rdar://problem/54660287>

Reviewed by Antti Koivisto.

Add bool m_isAnonymous member to mark anonymous layout boxes. Anonymous boxes are not rare enough to include this variable in RareData.

• layout/layouttree/LayoutBox.h:

(WebCore::Layout::Box::setIsAnonymous):

• layout/layouttree/LayoutTreeBuilder.cpp:

(WebCore::Layout::TreeBuilder::createLayoutBox):

6:31 AM Changeset in webkit [249084] by Antti Koivisto

• 4 edits
  2 adds in trunk/Source/WebCore

Implement layout system independent text box iterator
​https://bugs.webkit.org/show_bug.cgi?id=201076

Reviewed by Zalan Bujtas.

Add a generic way to traverse line layout without caring about the details of the underlying layout system.

This patch adds basic support for traversing text boxes and uses it to remove layout path specific branches in RenderTreeAsText.

• Sources.txt:
• WebCore.xcodeproj/project.pbxproj:
• rendering/RenderTreeAsText.cpp:

(WebCore::RenderTreeAsText::writeRenderObject):
(WebCore::writeTextBox):
(WebCore::write):
(WebCore::writeTextRun): Deleted.
(WebCore::writeSimpleLine): Deleted.

• rendering/line/LineLayoutInterfaceTextBoxes.cpp: Added.

(WebCore::LineLayoutInterface::TextBox::rect const):
(WebCore::LineLayoutInterface::TextBox::logicalRect const):
(WebCore::LineLayoutInterface::TextBox::hasHyphen const):
(WebCore::LineLayoutInterface::TextBox::isLeftToRightDirection const):
(WebCore::LineLayoutInterface::TextBox::dirOverride const):
(WebCore::LineLayoutInterface::TextBox::text const):
(WebCore::LineLayoutInterface::TextBoxIterator::TextBoxIterator):
(WebCore::LineLayoutInterface::TextBoxIterator::traverseNext):
(WebCore::LineLayoutInterface::TextBoxIterator::operator== const):
(WebCore::LineLayoutInterface::simpleLineRunResolverForText):
(WebCore::LineLayoutInterface::rangeForText):
(WebCore::LineLayoutInterface::TextBoxRange::TextBoxRange):
(WebCore::LineLayoutInterface::textBoxes):

• rendering/line/LineLayoutInterfaceTextBoxes.h: Added.

(WebCore::LineLayoutInterface::TextBox::TextBox):
(WebCore::LineLayoutInterface::TextBoxIterator::operator++):
(WebCore::LineLayoutInterface::TextBoxIterator::operator!= const):
(WebCore::LineLayoutInterface::TextBoxIterator::operator* const):
(WebCore::LineLayoutInterface::TextBoxRange::begin const):
(WebCore::LineLayoutInterface::TextBoxRange::end const):

6:24 AM Changeset in webkit [249083] by Alan Bujtas

• 2 edits in trunk/Source/WebCore

[LFC][BFC] Non-inline formatting context(table, grid etc) root container should not collapse through.
​https://bugs.webkit.org/show_bug.cgi?id=201099
<rdar://problem/54658946>

Reviewed by Antti Koivisto.

I didn't manage to find it in the spec, but surely formatting contexts like table, grid and flex should behave like
block so that their vertical margins are not adjoining, when they have at least one inflow child.

• layout/blockformatting/BlockMarginCollapse.cpp:

(WebCore::Layout::BlockFormattingContext::MarginCollapse::marginsCollapseThrough):

6:17 AM Changeset in webkit [249082] by Alan Bujtas

• 2 edits in trunk/Source/WebCore

[LFC] Add THEAD/TBODY/TFOOT output to Layout::showLayoutTree
​https://bugs.webkit.org/show_bug.cgi?id=201113
<rdar://problem/54664134>

Reviewed by Antti Koivisto.

• layout/layouttree/LayoutTreeBuilder.cpp:

(WebCore::Layout::outputLayoutBox):

6:16 AM Changeset in webkit [249081] by Alan Bujtas

• 4 edits in trunk/Source/WebCore

[LFC] Remove redundant Layout::Box::ElementType::TableRowGroup
​https://bugs.webkit.org/show_bug.cgi?id=201112
<rdar://problem/54663833>

Reviewed by Antti Koivisto.

Use Layout::Box::ElementType::TableBodyGroup instead.

• layout/layouttree/LayoutBox.cpp:

(WebCore::Layout::Box::isPaddingApplicable const):

• layout/layouttree/LayoutBox.h:
• layout/layouttree/LayoutTreeBuilder.cpp:

(WebCore::Layout::TreeBuilder::createLayoutBox):