We generated `PaintableFragment`s with a start and length represented in
UTF-8 byte offsets, but failed to consider that the offsets in a
`DOM::Range` are actually expressed in UTF-16 code units.
This is a bit of a mess: almost all web specs use UTF-16 code units as
the unit for indexing into text nodes, but we almost exclusively use
UTF-8 in our code base. Arguably the best thing would for us to use
UTF-16 everywhere as well: it prevents these mismatches in our
implementations for the price of a bit more memory usage - and even that
could potentially be optimized for.
But for now, try to do the correct thing and lazily allocate UTF-16 data
in a `PaintableFragment` whenever we need to index into it or if we're
asked to determine the code unit offset of a pixel position.
When clicking on a glyph or starting a selection on it, we would use the
glyph's offset/index as the position which represents the left side of
the glyph, or the position between the glyph and the glyph before it.
Instead of looking at which glyph is under the mouse pointer, look at
which glyph boundary is closer. Now, if you click to the right of a
glyph (but still on that glyph), it correctly selects the next glyph's
offset as the position.
We were passing in byte offsets instead of UTF-16 code unit offsets,
which could lead to crashes if the offsets found exceeded the number of
code units in text fragments on the page.
Fixes#4908.
Co-authored-by: Tim Ledbetter <tim.ledbetter@ladybird.org>
The text track processing model would previously spin forever waiting
for the track URL to change. It would then recursively invoke itself
to handle the new URL, again entering the spin loop. This meant that
tracks could easily cause event loop hangs.
We now have an observer system to be notified when the track state
changes instead. This lets us exit the processing model and move on.
We have a test for localStorage that repeatedly adds new items until the
quota is exceeded:
`webstorage/storage_local_setitem_quotaexceedederr.window.html`.
This test becomes significantly slower when localStorage is backed by
SQL database. Let's disable database usage in test mode for now, as this
test is likely to be flaky on CI due to timeouts.
The "longhands" array is populated in the code generator to avoid the
overhead of manually maintaining the list in Properties.json
There is one subtest that still fails in
'cssstyledeclaration-csstext-all-shorthand', this is related to
us not maintaining the relative order of CSS declarations for custom vs
non-custom properties.
url() has some limitations because of allowing unquoted URLs as its
contents. For example, it can't use `var()`. To get around this, there's
an alternative `src()` function which behaves the same as `url()` except
that it is parsed as a regular function, which makes `var()` and friends
work properly.
There's no WPT test for this as far as I can tell, so I added our own.
You would have to just know that you need to define the constructor with
this declaration. Let's allow subclasses to define constructors as they
see fit.
This is causing errors on the WPT runner, which does not have a display
output. To do this requires shuffling around the Main::Arguments struct,
as we now need access to it from overridden WebView::Application methods
after construction.
Now that headless mode is built into the main Ladybird executable, the
headless-browser's only purpose is to run tests. So let's move it to the
testing directory and rename it to test-web (a la test-js / test-wasm).
Removes the associated FIXME in match_an_audio_or_video_type_pattern().
Sniffing process is a simplified version of the full spec, as it only
checks one frame of the mp3. To fully align with the spec, it would
also have to check a second frame by calculating frame size as
described in the spec.
Some instances of CSSStyleProperties can lack an owner node, for
instance the return value of a call to `window.getComputedStyle` where
the specified pseudo-element is invalid. In this case we should treat
the computed style as empty, as there is no node to compute the style
for.
Previously we would just throw it away and construct a new (empty) one
when required. This doesn't work as any existing references to the old
instance will contain out of date information. Now we retain and update
the existing instance instead.
Resolves a FIXME in `CSSRuleList::insert_a_css_rule`. Gets us a bit
closer to passing https://wpt.live/css/cssom/at-namespace.html but that
requires more work around parsing of selectors with namespaces (namely
disallowing use of undeclared selectors), which I have added a FIXME
for.
The spec requires us to store properties in their shorthand-expanded
form and in the "specified" order, removing duplicates prefering based
on "cascading order". We already enforced this in `set_property` but
not at creation time (e.g. in stylesheets) or in `set_css_text` (e.g.
updating style attribute).
This commit enforces the spec requirements at all the relevant points.
We no longer include logical properties in the return value of
`getComputedStyle` as they are mapped to their physical equivalents in
`StyleComputer::for_each_property_expanding_shorthands`, but resolving
that requires a relatively large rework of how we handle logical
properties, (namely moving when we map them to their physical
equivalents from parse time to style computation time).
This also exposes two false positive tests in
wpt-import/css/cssom/border-shorthand-serialization.html related to us
not yet supporting the border-image css property.
We assume elsewhere that any abspos element's containing block must be
some kind of Layout::Box, so let's enforce that when deciding if a box
can be such a container.
This fixes a bad downcast on https://serpapi.com/
An early return was occurring between the emission of
PushStackingContext and PopStackingContext, resulting in a
PushStackingContext without a corresponding PopStackingContext in the
display list, which caused broken painting.
Fixes black screen on Discord login page.
`BlockFormattingContext::compute_width()` stores the left and right
margins in the layout state at the very end of the function. However,
before doing so, it calls `FormattingContext::calculate_inner_width()`
which ends up calling `FormattingContext::calculate_stretch_fit_width()`
if the current box has `width: fit-content`.
Due to this, `calculate_stretch_fit_width()` would always see the
margins from the layout state as zero and therefore not take them into
account. Subsequently, the calculated width ended up being wrong.
Saving margins on the layout state earlier, before calling
`calculate_inner_width()`, makes sure that the width is calculated
correctly.
An earlier variant of the commit following this one introduced a
regression for the behavior tested here, but did not fail any in-tree
tests. So lets add an explicit regression test to make that easier to
catch in the future.
Based very scientifically on what's listed here:
https://harfbuzz.github.io/what-does-harfbuzz-do.html
I've moved the code into LibGfx because including a HarfBuzz header
directly from LibWeb is a little unpleasant. But the Gfx::FontTech enum
follows the CSS definitions for font features for simplicity.
TrueType collections are supported. SVG and Embedded OpenType are not,
but they're not widely supported by other browsers so that's fine.
Most of the features are completely supported by HarfBuzz, so we can
just return true. Graphite support is optional (and it appears we use a
build of HarfBuzz without it) but there's a define we can check.
Incremental Font Transfer is a whole separate thing that we definitely
don't support yet.
We previously checked the cell's computed values for the border-collapse
property, but a cell is only in separated-borders mode if the table has
border-collapse set to separate. Curiously in some other placed where
this mode is checked we already did the correct thing.
For getComputedStyle(), we must return an absolute URL for image style
values. We currently return the raw parsed URL.
This fixes loading the marker icons on https://usermap.serenityos.org.
When marking a part of the layout tree for rebuild, if the subtree root
that we're marking has an anonymous parent, we now mark from the nearest
non-anonymous ancestor instead.
This ensures that subtrees inside anonymous wrappers don't just get
duplicated (i.e recreated but inserted instead of replaced).
The test CSSStyleDeclaration-has-indexed-property-getter is a frequent
source of merge conflicts between PRs that are adding to or otherwise
modifying the list of supported CSS properties.
This is primarily because the test prints out a numeric index of each
property along with the property. As far as I can tell the indexes are
inconsequential for what the test is trying to verify. So lets modify
the printout to only contain the properties without indexes.