Bug 99228

I found that we have to change the current algorithm to reproject <shadow>. Currently, we resolve <content> from the youngest shadow root. However, when we reproject <shadow>, we have to resolve <content> and <shadow> from the oldest shadow root, since we have to know what elements are distributed to <shadow>. This is an example. host ------- SR -------------- SR (younger) |- div (A) |-div (C) |-div -------- SR |- div (B) |-content[B] |-shadow |-content[div]#target |-shadow When we resolve content[div]#target, we have to resolve all the other shadow and content. This might be a drastic spec change.

I think we can punt on this for now, possibly forever.

(In reply to comment #2) > I think we can punt on this for now, possibly forever. When we adopt OOP model to understand reprojection, reprojecting shadow seems like seeing a private field of a base class. (reprojecting content is just passing arguments to super class.) I think it would be better to postpone implementing this until we strongly require this.

I'm now thinking implementing this is not so hard...

Created attachment 170094 [details] WIP

This WIP includes code related to https://bugs.webkit.org/show_bug.cgi?id=99552 also. I've confirmed that distribution works, but we have to fix ComposedShadowTreeWalker also.

Comment on attachment 170094 [details] WIP This looks promising!

Created attachment 170338 [details] WIP

Now shadow-reprojection works in some case, though we have a lot of failing tests.

<shadow> in <shadow> does not work again :-/

Created attachment 170352 [details] WIP

The last WIP patch passes all the existing fast/dom/shadow tests. I'll add more tests and refine the code tomorrow.

Created attachment 170527 [details] WIP

Now writing a composed shadow tree walker test, and it's discovering a lot of failures. I need more time to make it review-ready.

We have to update the AncestorChainWalker also. AncestorChainWalker::parent() if (!m_node->isShadowRoot()) { m_node = m_node->parentNode(); (*) m_distributedNode = m_node; m_isCrossingInsertionPoint = false; return; } (*) Here, we should preserve m_node as a m_distributed node before assigning m_node->parentNode() to m_node if the following condition is met: m_node->parenetNode()->isShadowRoot() && toShadowRoot(m_node->parentNode())->isAssignedTo() since m_node might be re-projected. So we should keep it as a m_distributed_node. Let me take a look at other places also. I am afraid that we have to update other places. (In reply to comment #14) > Now writing a composed shadow tree walker test, and it's discovering a lot of failures. I need more time to make it review-ready.

Created attachment 170566 [details] WIP

I've taken a look at EventDispatcher::ensureEventAncestors. I expect that we don't have to update EventDispatcher as long as we can calculate ancestors chain in AncestorChainWalker correctly. (In reply to comment #15) > We have to update the AncestorChainWalker also. > > AncestorChainWalker::parent() > > if (!m_node->isShadowRoot()) { > m_node = m_node->parentNode(); (*) > m_distributedNode = m_node; > m_isCrossingInsertionPoint = false; > return; > } > > (*) Here, we should preserve m_node as a m_distributed node before assigning m_node->parentNode() to m_node if the following condition is met: > > m_node->parenetNode()->isShadowRoot() && toShadowRoot(m_node->parentNode())->isAssignedTo() > > since m_node might be re-projected. So we should keep it as a m_distributed_node. > > Let me take a look at other places also. I am afraid that we have to update other places. > > > > (In reply to comment #14) > > Now writing a composed shadow tree walker test, and it's discovering a lot of failures. I need more time to make it review-ready.

> We have to update the AncestorChainWalker also. > Done. What we have to fix now is: - ComposedShadowTreeWalker::nextSibling(), previousSibling(). If you find other places where we have to fix, please tell me.

Comment on attachment 170566 [details] WIP Attachment 170566 [details] did not pass qt-ews (qt): Output: http://queues.webkit.org/results/14593126

Comment on attachment 170566 [details] WIP Attachment 170566 [details] did not pass qt-wk2-ews (qt): Output: http://queues.webkit.org/results/14555512

Comment on attachment 170566 [details] WIP Attachment 170566 [details] did not pass efl-ews (efl): Output: http://queues.webkit.org/results/14551008

Comment on attachment 170566 [details] WIP Attachment 170566 [details] did not pass chromium-ews (chromium-xvfb): Output: http://queues.webkit.org/results/14566506

Comment on attachment 170566 [details] WIP Attachment 170566 [details] did not pass cr-android-ews (chromium-android): Output: http://queues.webkit.org/results/14562497

Comment on attachment 170566 [details] WIP View in context: https://bugs.webkit.org/attachment.cgi?id=170566&action=review > Source/WebCore/dom/ComposedShadowTreeWalker.cpp:67 > +static inline InsertionPoint* resolveReprojection(const Node* projectedNode, const Node* baseNode) Do we need two parameters? it seems one parameter, projectedNode, is enough. > Source/WebCore/dom/ComposedShadowTreeWalker.cpp:86 > + return resolveShadowProjection(root->assignedTo()); I am afraid that this does not resolve reprojection for shadow. If we resolve reprojection recursively, we need two nodes as state, one is an insertion point and the other is a re-projected node. > Source/WebCore/dom/ComposedShadowTreeWalker.cpp:223 > + InsertionPoint* insertionPoint = resolveReprojection(node, node); One parameter seems to be enough. > Source/WebCore/dom/ComposedShadowTreeWalker.cpp:291 > + if (InsertionPoint* insertionPoint = resolveReprojection(node, node)) { Ditto.

Comment on attachment 170566 [details] WIP Attachment 170566 [details] did not pass mac-ews (mac): Output: http://queues.webkit.org/results/14554656

Created attachment 170766 [details] WIP

It started to pass the tests again.

Comment on attachment 170766 [details] WIP View in context: https://bugs.webkit.org/attachment.cgi?id=170766&action=review > Source/WebCore/css/StyleResolver.cpp:1547 > + } memo: This is necessary because the direct child of shadow root can be reprojected to somewhere now. In that case, parentElement() will return 0. > Source/WebCore/dom/ComposedShadowTreeWalker.cpp:108 > + This should die.

Created attachment 170782 [details] WIP

This version works correctly including ComposedShadowTreeWalker. I'll refine the code.

Created attachment 170788 [details] Patch

Created attachment 170792 [details] Patch

Fixed ChangeLog mojibake

(In reply to comment #33) > Fixed ChangeLog mojibake Sounds delicious.

Created attachment 170801 [details] Patch

Minor updates: removed a debug code, and ChangeLog update.

Created attachment 170807 [details] Patch

Added a testcase about inert InsertionPoint, and fixed a bug related to it.

Hmm... No one did review? :-(

(In reply to comment #38) > Added a testcase about inert InsertionPoint, and fixed a bug related to it. I'll try to get this in a couple of hours. Sorry :-\

Comment on attachment 170807 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=170807&action=review > Source/WebCore/dom/ElementShadow.cpp:135 > InsertionPoint* ElementShadow::insertionPointFor(const Node* node) const That method got whittled away to nothing! Can probably remove it. > Source/WebCore/html/shadow/InsertionPoint.h:135 > +inline Element* parentElementForDistribution(const Node* node) Is this guy still being used somewhere?

(In reply to comment #41) > (From update of attachment 170807 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=170807&action=review > > > Source/WebCore/dom/ElementShadow.cpp:135 > > InsertionPoint* ElementShadow::insertionPointFor(const Node* node) const > > That method got whittled away to nothing! Can probably remove it. > > > Source/WebCore/html/shadow/InsertionPoint.h:135 > > +inline Element* parentElementForDistribution(const Node* node) > > Is this guy still being used somewhere? Thanks for reviewing! I'll make follow-up patches after landing this.

Comment on attachment 170807 [details] Patch Clearing flags on attachment: 170807 Committed r132760: <http://trac.webkit.org/changeset/132760>

All reviewed patches have been landed. Closing bug.