Emitting tokens on EOF caused an infinite loop, freezing the app, which
could be a bit annoying when writing an HTML comment at the end of
the file in Text Editor. :^)
Commit b193351a99 caused the HTML comments to flash when changing
the text cursor. Also, when double-clicking on a comment, the selection
started from the beginning of the file instead.
The following message was displaying when `TOKENIZER_TRACE_DEBUG`
was enabled:
(Tokenizer::nth_last_position) Invalid position requested: 4th-last
of 4. Returning (0-0).
Changing the `nth_last_position` to 3 fixes this. I'm guessing that's
because the parser is at that moment on the second hyphen of the `<!--`
string, so it has to go back only by three characters.
This reverts commit 3a184f7841.
This broke a number of test262 tests under "TypedArrayConstructors".
The issue is that the CanonicalNumericIndexString AO should not fail
for inputs like "1.1", despite them not being integral indices.
The difference should be between m_utf8_iterator and the
the new position, if m_prev_utf8_iterator is used one fewer
source position is popped than required.
This issue was not apparent on most pages since restore_to
used for tokens such <!doctype> that are normally
followed by a newline that resets the column to zero,
but it can be seen on pages with minified HTML.
The spec version of canonical_numeric_index_string is absurdly complex,
and ends up converting from a string to a number, and then back again
which is both slow and also requires a few allocations and a string
compare.
Instead lets use the logic we already have as that is much more
efficient.
This improves performance of all non-numeric property names.
getClientRects supposed to return a list of bounding DOMRect
for each box fragment of Element's layout, but most elements have
only one box fragment, so implementing it with getBoundingClientRect
is useful.
Until now, some formatting contexts (BFC in particular) have been
assigning size to the root box. This is really the responsibility of the
parent formatting context, so let's stop doing it.
To keep position:absolute working, parent formatting contexts now notify
child contexts when the child's root box has been sized. (Note that the
important thing here is for the child root to have its final used height
before it's able to place bottom-relative boxes.)
This breaks flexbox layout in some ways, but we'll have to address those
by improving the spec compliance of FFC.)
We sometimes had a stale stacking context tree sitting around, causing
incorrect paints until the next full layout invalidation.
Fix this by simply rebuilding the stacking context tree when asked to.
BFC currently has a number of architectural issues due to it being
responsible for setting the dimensions of the BFC root.
This patch moves the logic for setting up the ICB from BFC to Document.
Since we were previously relying on Document::set_cookie in order to
set cookies received as a 'Set-Cookie' response header, we would ignore
any response header cookies in redirect (status code 3xx) responses.
While this behaviour is not strictly enforced in the specification,
most major browsers do set cookies in redirect responses, and some
sites (e.g. Cookie Clicker) rely on this behaviour.
Since cookies are stored per-site and not per-document, this behaviour
is achieved by simply decoupling the cookie set mechanism from it.
As per RFC7231 the Location header field has different meanings for
different response status codes:
For 201 (Created) responses, the Location value refers to the primary
resource created by the request.
For 3xx (Redirection) responses, the Location value refers to the
preferred target resource for automatically redirecting the request.
The MIME type can come from either the Content-Type header in the
response header list, or the overridden MIME type passed into
overrideMimeType()
This is required by Google Maps, which downloads blobs encoded
in x-user-defined format.
This allows you to ignore the Content-Type returned by the server and
always parse the content as if it's the given MIME type.
This will currently be used for allowing you to override the charset
of text responses.
This is all still quite ad-hoc. Eventually these will both need to
support units (like with CSS Lengths) but for now we can continue only
using numbers.
I've chosen the name `AttributeParser` since it parses data from
attributes. Rather than duplicate the parsing of numbers and other
basic types, let's make use of this existing parsing code for parsing
the data for `<line>`, `<polyline>`, etc.
From the spec:
> Interface SVGGeometryElement represents SVG elements whose rendering
> is defined by geometry with an equivalent path, and which can be
> filled and stroked. This includes paths and the basic shapes.
- https://svgwg.org/svg2-draft/types.html#InterfaceSVGGeometryElement
Making them all create an SVGPathBox, and return a Path from get_path(),
means we can implement the "basic shapes" using the path system we
already have. :^)
This patch introduces the StyleComputer::RuleCache, which divides all of
our (author) CSS rules into buckets.
Currently, there are two buckets:
- Rules where a specific class must be present.
- All other rules.
This allows us to check a significantly smaller set of rules for each
element, since we can skip over any rule that requires a class attribute
not present on the element.
This takes the typical numer of rules tested per element on Discord from
~16000 to ~550. :^)
We can definitely improve the cache invalidation. It currently happens
too often due to media queries. And we also need to make sure we
invalidate when mutating style through CSSOM APIs.
Previously we would re-run the entire CSS selector machinery for each
property resolved. Instead of doing that, we now resolve a final set of
custom property key/value pairs at the start of the cascade.
- Replace "auto" with "auto const" where appropriate.
- Remove an unused struct.
- Make sort_matching_rules() a file-local static function.
- Remove some unnecessary includes.
Our previous code roughly did this:
1. Generate a bitmap as large as the shadow would end up.
2. Paint a rectangle onto it.
3. Blur the whole bitmap.
4. Split it up and render each section.
This patch takes advantage of the fact that (aside from corners) each
horizontal or vertical strip of a box-shadow is identical to the
others, to generate and blur a much smaller bitmap - only large enough
for the four corners and 1px of central "side" in each direction. This
greatly reduces the memory footprint, and should also speed things up,
since there is much less to blur.