A challenge for getting LibTest working on Windows has always
been CrashTest. It implements death tests similar to Google Test
where a child process is cloned to invoke the expression that
should abort/terminate the program. Then the exit code of the
child is used by the parent test process to verify if the
application correctly aborted/terminated due to invoking
the expression.
The problem was that finding an equivalent way to port Crash::run()
to Windows was not looking very likely as publicly exposed Win32/
Native APIs have no equivalent to fork(); however, Windows actually
does have native support for process cloning via undocumented NT
APIs that clever people reverse engineered and published, see
`NtCreateUserProcess()`.
All that being said, this `EXPECT_DEATH()` implementation avoids
needing to use a child process in general, allowing us to remove
CrashTest in favour of a single cross-platform solution for death
tests.
When a `VERIFY()` assertion fails `ak_verification_failed` is invoked
which means the program will invoke `__builtin_trap` and immediately
quit. But with this change we have now allowed for an optional hook
into `ak_*_failed` that lets us perform some custom
action before program exits. This foundation is what we will
(ab)use to implement death tests without the process cloning
CrashTest infrastructure.
Originally part of a fix in 15103d172c, it
appears that this is no longer necessary and received a better fix in a
more recent commit. Resolves a visual regression with the ACID3 test.
The main motivation here is that the CSS Parser needs to know about
PageSelectorList so that we can parse one in
`CSSPageRule::set_selector_text()`. Including all of `CSSPageRule.h`
there would pull in a lot of other headers that aren't needed.
Previously we only matched against the first attribute with a given
local name. What we actually want to do is look at each attribute with
that local name in turn and only return false if none of them match.
Also remove a hack for HTML elements in HTML documents, where we would
refuse to match any namespaced attributes. This doesn't seem to be
based on the spec, but we had regressions without it, until now. :^)
Gets us 21 more WPT subtest passes.
The HTML spec gives us a list of HTML attributes that must have their
values compared case-insensitively by default (when the attribute
selector does not specify a case-sensitiveness). However, ifwe have a
namespace, then we are not looking for an HTML attribute, so this
should not apply.
Gets us 8 more WPT subtest passes.
This is a bit under-specced, specifically there's no definition of
CSSMarginDescriptors so I've gone with CSSStyleProperties for now. Gets
us 17 WPT subtests.
By doing that we avoid lots of `PropertyKey` -> `Value` -> `PropertyKey`
transforms, which are quite expensive because of underlying
`FlyString` -> `PrimitiveString` -> `FlyString` conversions.
10% improvement on MicroBench/object-keys.js
With this, we pass the 8 ref tests in css/selectors/selectors-4/ which
previously failed. This is not technically a full implementation, as we
are supposed to first canonicalize the language range and tag, but that
will require downloading and processing the IANA language subtag
registry:
https://www.iana.org/assignments/language-subtag-registry/language-subtag-registry
That's significantly more work, and WPT doesn't seem to test any cases
that require that, so we can leave it for now.
Adds `matches_webm_signature()` and `parse_vint()` helpers per WPT
spec. Uses these helpers to resolve the WebM FIXME that was in
`match_an_audio_or_video_type_pattern()`.
This commit adds support in AK/Random for a high quality RNG on windows.
This requires moving the code into a cpp file not to spread windows
headers around.
This commit adds a convenience method to secure random for initializing
single types. It changes the random number generator in JS math random()
to use newer constants by the author as well as initializes it with a
higher quality seed.
For our js-benchmarks and libjs-test262 workflow runs, we already know
that they're provisioned with these repositories and can skip adding the
key and repo altogether.
We were using both wget and curl arbitrarily; use curl exclusively since
that is installed by default on our machines and containers. Fixes the
js-benchmarks workflow.
This was previously used by the WPT runner to determine the git hash
of a particular WPT run. This mechanism is no longer used, since it
doesn't work with chunked WPT runs.
Our Ranges should maintain the invariant that their offsets are always
within range of 0..length (inclusive) of their respective containers.
Note that we cannot maintain this in AbstractRange, which is the base
for StaticRange and can still have invalid offsets.
156c1083e9 introduced a text blocks cache
for better performance when searching through text on a page, but when
we partially recreate the layout tree, this cache does not get
invalidated. We now rebuild the entire text blocks cache after a layout
update.
We were calling into `Range::set_start_or_end()` indirectly through
`::set_start()` and `::set_end()`, but that algorithm only calls for an
invocation whenever the start or end of a range needs to be set to a
boundary point. If an algorithm step calls for setting the node or
offset, we should directly modify the range.
The problem with calling into `::set_start_or_end()` is that this
algorithm potentially modifies _both_ the start and end of the range,
but algorithms trying to update a range's start or end often have
explicit steps to take both the start and end into account and end up
overcompensating for the start or end offset resulting in an invalid
range (e.g. with an end offset beyond a node's length).
This makes updating a range's start/end a bit more efficient and removes
a piece of ad-hoc code in CharacterData needed to make it work before.
I was wrong when I added those notes before about this being impossible,
it's *very* possible, for example with the `@page margin` descriptor.
However, until we have a large number of these shorthands and not just a
single example, we can get away with hard-coding support for it.
Ideally we'd be able to share the code between page selectors and style
ones, but given how simple page selectors are, some code duplication is
the simpler option.
