Problem: Discarding a document that contains an SVG “use” element could
abort the process with a !is_in_list() verification failure in the
IntrusiveListNode destructor. That surfaced intermittently in our style-
invalidation stress tests, depending on GC sweep order.
Cause: A “use” element connected to a document registers itself in the
document’s list of “use” elements and unregisters during its removal
steps. A GC’ed “use” element is swept without running those removal
steps — so it stays linked. When it’s destroyed before its document,
its list node is still linked — and the destructor aborts.
Fix: Override finalize() to unregister the “use” element before
destruction. The collector finalizes every dying cell before destroying
any of them. So, the node is always unlinked in time — the same approach
DocumentObserver and NavigationObserver already use.
Every SVG element insertion, removal, attribute change, and children
change walked the entire document looking for use elements to notify
about possible referenced-subtree changes. On pages with large SVG
documents this is quadratic: loading chatgpt.com spent 7% of all CPU
samples in these full-document scans, nearly all of it while parsing
an SVG icon sprite sheet.
Instead, keep every use element connected to a document's node tree in
an intrusive list owned by that document, and only iterate that list
(usually empty or tiny) when an SVG element changes.
Subtleties:
- A use element inserted by the same subtree insertion as its
referenced element, but after it in tree order, used to be found by
the document-wide scan from the referenced element's insertion
steps. Now SVGUseElement::inserted() re-resolves the reference if
the shadow tree is still unpopulated. A new test covers both tree
orders.
- Node.moveBefore() runs moving steps without insertion or removal
hooks. Now SVGUseElement::moved_from() updates list membership when
moving across document-tree and shadow-tree boundaries. A new test
covers both directions.
- Removal hooks run after the subtree has been detached, so use
elements being removed alongside the changed element may still be
registered. Filter them out structurally via root().is_document(),
since Node::is_connected() is a flag that is updated in hook order
and can still be stale at this point.
Otherwise, the load event will block the original document until GC
runs.
Without this, media-load-task-after-adoption.html would wait for the
idle timeout to trigger a garbage collection, which could sometimes
cause the test to time out entirely.
Move the layout tree from GC allocation to refcounted ownership so
removed layout and paint subtrees are destroyed synchronously instead
of waiting for the next GC sweep. This dramatically reduces GC memory
usage peaks after layout tree churn and makes it easier for memory use
to fall back after large document updates.
Update layout factories, tree traversal, SVG layout node creation,
paintable back-pointers, and pseudo-element layout links to use RefPtr
ownership.
Make display: contents follow the same shape as Blink and WebKit: the
element itself does not create a layout node, and its children are
flattened into the nearest layout parent. Wrap direct non-whitespace
text in an anonymous inline node when the boxless element contributes
inherited style to that text.
Use an internal inline wrapper for display: contents pseudo-elements
so generated content can still participate in layout, painting, hit
testing, and pseudo-element queries. Keep CSSOM reporting the computed
display value from the pseudo style, not the internal wrapper.
Remove the retained out-of-tree layout node list and its testing hook,
since the flattened model does not need a side owner for boxless
elements. Add coverage for inherited text style, dynamic insertion
order, pseudo-element hit testing, and computed style queries.
Move ComputedProperties and CascadedProperties out of the GC. They no
longer contain strong references to GC-managed data.
Keep computed styles alive from DOM elements and animation updates with
RefPtr. Pass style into layout constructors by reference, since layout
only copies the values it needs while building nodes.
Use GC::Weak for cascade source links, so entries no longer keep the
style declaration or shadow root alive.
Skip the document-complete clone when a resolved SVG use element already
has a current shadow tree. Href processing creates that root for refs.
Replacing it during load can leave stale mask contents.
Still clone at document-complete for forward refs and changed refs.
Add coverage for SVG use in masks, pre-load subtree mutations, and
unchanged instance roots.
Previously, the LibWeb bindings generator would output multiple per
interface files like Prototype/Constructor/Namespace/GlobalMixin
depending on the contents of that IDL file.
This complicates the build system as it means that it does not know
what files will be generated without knowledge of the contents of that
IDL file.
Instead, for each IDL file only generate a single Bindings/<IDLFile>.h
and Bindings/<IDLFile>.cpp.
Previously, referencing an element that eventually pointed back to
the original <use> element would cause a stack overflow.
We now look at the referenced element and follow every <use> element
in its subtree and aim to detect any duplicates. If so, we consider
that referenced element invalid.
`SVGUseElement::referenced_element()` previously passed
`m_href->fragment()` directly to `Document::get_element_by_id()`. URL
fragments are stored in their serialized form, so references to elements
whose IDs contain non-ASCII characters would fail to resolve.
Prevents observably calling Trusted Types, which can run arbitrary JS,
cause crashes due to use of MUST and allow arbitrary JS to modify
internal elements.
An SVGLength can be read-only, e.g. all animVal values cannot be
modified. Implement this for all instantiations of SVGLength.
While we're here, add `fake_animated_length_fixme()` so we can easily
find all sites where we need to improve our animated length game.
Before this change, we were going through the chain of base classes for
each IDL interface object and having them set the prototype to their
prototype.
Instead of doing that, reorder things so that we set the right prototype
immediately in Foo::initialize(), and then don't bother in all the base
class overrides.
This knocks off a ~1% profile item on Speedometer 3.
This full invalidation was just papering over earlier bugs in the
partial layout tree update code. The DOM mutations that happen here
should be enough to drive the necessary invalidation now.
Note that this is covered by a regression test added with the
invalidation.
This is sub-optimal but let's rebuild the whole tree for now, since this
case gets quite complicated and there are more valuable things to chase
after first.
Thanks to Gingeh for the reduced test case!
Resulting in a massive rename across almost everywhere! Alongside the
namespace change, we now have the following names:
* JS::NonnullGCPtr -> GC::Ref
* JS::GCPtr -> GC::Ptr
* JS::HeapFunction -> GC::Function
* JS::CellImpl -> GC::Cell
* JS::Handle -> GC::Root
We currently have 2 virtual methods to inform DOM::Element subclasses
when an attribute has changed, one of which is spec-compliant. This
patch removes the non-compliant variant.
Now that the heap has no knowledge about a JavaScript realm and is
purely for managing the memory of the heap, it does not make sense
to name this function to say that it is a non-realm variant.
The main motivation behind this is to remove JS specifics of the Realm
from the implementation of the Heap.
As a side effect of this change, this is a bit nicer to read than the
previous approach, and in my opinion, also makes it a little more clear
that this method is specific to a JavaScript Realm.
