There's a chance that we try to choose a navigable before a previously
destroyed navigable is fully destroyed and GC'd. Investigating why this
can happen is a separate endeavor, let's just not crash for now.
This allows for:
* Transformed text (e.g. rotated text)
* Stroked text
* Filling/stroking text with PaintStyles (e.g. gradients)
* Squashed/condensed text (via maxWidth parameter)
Fixes part of #22817
Previously, 'now' was set to the time `requestAnimationFrame()` was
called, and the EventLoop's 'now' was ignored. This was a little odd and
meant the time was always in the past.
In this commit we have optimized the handling of scroll offsets and
clip rectangles to improve performance. Previously, the process
involved multiple full traversals of the paintable tree before each
repaint, which was highly inefficient, especially on pages with a
large number of paintables. The steps were:
1. Traverse the paintable tree to identify all boxes with scrollable or
clipped overflow.
2. Gather the accumulated scroll offset or clip rectangle for each box.
3. Perform another traversal to apply the corresponding scroll offset
and clip rectangle to each paintable.
To address this, we've adopted a new strategy that separates the
assignment of the scroll/clip frame from the refresh of accumulated
scroll offsets and clip rectangles, thus reducing the workload:
1. Post-relayout: Identify all boxes with overflow and link each
paintable to the state of its containing scroll/clip frame.
2. Pre-repaint: Update the clip rectangle and scroll offset only in the
previously identified boxes.
This adjustment ensures that the costly tree traversals are only
necessary after a relayout, substantially decreasing the amount of work
required before each repaint.
When WebDriver asks to destroy a window, we can hit this case with no
active browsing context. This seems odd, but perhaps is a spec issue as
well. Just log to dbgln for now.
This aligns it better with the current state of the spec.
There's still some functions and data members that need moved into
Navigable or TraversableNavigable, but we can leave those for the next
cleanup PR.
The IPC layer between chromes and LibWeb now understands that multiple
top level traversables can live in each WebContent process.
This largely mechanical change adds a billion page_id/page_index
arguments to make sure that pages that end up opening new WebViews
through mechanisms like window.open() still work properly with those
extra windows.
When an element with an ID is added to or removed from the DOM, or if
an ID is added, removed, or changed, then we must reset the form owner
of all form-associated elements who have a form attribute.
We do this in 2 steps, using the DOM document as the messenger to handle
these changes:
1. All form-associated elements with a form attribute are stored on the
document. If the form attribute is removed, the element is removed
from that list as well.
2. When a DOM element with an ID undergoes any of the aforementioned
changes, it notifies the document of the change. The document then
forwards that change to the stored form-associated elements.
By replacing the `page_did_request_scroll_to()` calls with a request
to perform scrolling in the corresponding navigable, we ensure that
the scrolling of iframes will scroll within them instead of triggering
scroll of top level document.
Recently, we moved the resolution of CSS properties that do not affect
layout to occur within LayoutState::commit(). This decision was a
mistake as it breaks invalidation. With this change, we now re-resolve
all properties that do not affect layout before each repaint.
Paintable boxes should not hold information stored in device pixels.
It should be converted from CSS pixels only by the time painting
command recording occurs.
With this change, clip rectangles for boxes with hidden overflow or the
clip property are no longer calculated during the recording of painting
commands. Instead, it has moved to the "pre-paint" phase, along with
the assignment of scrolling offsets, and works in the following way:
1. The paintable tree is traversed to collect all paintable boxes that
have hidden overflow or use the CSS clip property. For each of these
boxes, the "final" clip rectangle is calculated by intersecting clip
rectangles in the containing block chain for a box.
2. The paintable tree is traversed another time, and a clip rectangle
is assigned for each paintable box contained by a node with hidden
overflow or the clip property.
This way, clipping becomes much easier during the painting commands
recording phase, as it only concerns the use of already assigned clip
rectangles. The same approach is applied to handle scrolling offsets.
Also, clip rectangle calculation is now implemented more correctly, as
we no longer stop at the stacking context boundary while intersecting
clip rectangles in the containing block chain.
Fixes:
https://github.com/SerenityOS/serenity/issues/22932https://github.com/SerenityOS/serenity/issues/22883https://github.com/SerenityOS/serenity/issues/22679https://github.com/SerenityOS/serenity/issues/22534
Also change a Vector<Handle> to a Vector<NonnullGCPtr> while we're
here, since there's no need to use handles for members of a cell.
Fixes an ASAN error on the HTML/Window-postMessage.html test.
This change makes WebGL to use LibGL only in SerenityOS, and the
platform's OpenGL driver in Ladybird if it is available.
This is implemented by introducing wrapper class between WebGL and
OpenGL calls. This way it will also be possible to provide more
complete support in Ladybird even if we don't yet have all needed
calls implemented in LibGL.
For now, the wrapper class makes all GL calls virtual. However, we
can get rid of this and implement it at compile time in case of
performance problems.
In the cases where spec authors have us directly interact with promises
in a task source context, we need to prepare the backup settings object
stack as well as push an actual execution context to the JS VM.
This prevents goal conditions that rely on tasks that are currently in
flight on the task queue (or were just submitted) from blocking the
event loop until an IPC event fires to kick the native event loop.
This involves passing the UserNavigationInvolvement from each form
associated element that triggers a submit through to the methods that
perform the actual navigation. While here, refactor HTMLFormElement to
use the new Bindings::NavigationHistoryBehavior enum.
The one current caller of this function always defers microtask
checkpoints before calling wait_for_all, ensuring that the promise
accept/reject handlers will always be called later in the Web event loop
processing. We need to store all the state for the closures in a heap
allocated object with HeapFunctions to keep it around while there are
still promises to resolve.
