We have a 5 second timeout between a user-activated event occurring and
an activation-gated API being invoked in order for that API to succeed.
This is quite fine in normal circumstances, but the machines used in CI
often exceed that limit (we see upwards of 10 seconds passing between
generating the user-activated event and the API call running).
So instead of generating a user-activated event, add a hook to allow
tests to bypass the very next activation check.
Instead of having Call refer to a range of VM registers, it now has
a trailing list of argument operands as part of the instruction.
This means we no longer have to shuffle every argument value into
a register before making a call, making bytecode smaller & faster. :^)
By handling common cases like Int32 arithmetic directly in the
instruction handler, we can avoid the cost of calling the generic helper
functions in Value.cpp.
Instead of splitting the postfix variants into ToNumeric + Inc/Dec,
we now have dedicated PostfixIncrement and PostfixDecrement instructions
that handle both outputs in one go.
If a call to `document.write` inserts an incomplete HTML tag, e.g.:
document.write("<p");
we would previously continue parsing the document until we reached a
closing angle bracket. However, the spec states we should stop once we
reach the new insertion point.
We were previously embedding the original text to handle the special
case where the text is empty. We generate an extra span to hold the
string "#text" as a placeholder, so that we don't generate a 0px-wide,
unclickable (and therefore uneditable) node. Instead, we can just detect
when this is the case in the Inspector JS.
This further reduces the generated HTML for the Inspector on
https://github.com/SerenityOS/serenity from 1.9MB to 1.8MB (about 94KB,
or 4.7%).
Attribute values may contain HTML, and may contain invalid HTML at that.
If the latter occurs, let's not generate invalid Inspector HTML when we
embed the attribute values as data attributes. Instead, cache the values
in the InspectorClient, and embed just a lookup index into the HTML.
This also nicely reduces the size of the generated HTML. The Inspector
on https://github.com/SerenityOS/serenity reduces from 2.3MB to 1.9MB
(about 318KB, or 13.8%).
We currently support optional integral types, but not nullable types. So
if an IDL contains e.g. "long?", passing null will be coerced to 0.
This will be used by the Inspector, but will also eventually be used by
real IDL interfaces (e.g. HTMLInputElement's selectionStart).
This is required by the CFF spec, and is consistent with what we do for
the encoding 24 lines down.
As far as I can tell, nothing in `Type1FontProgram::rasterize_glyph()`
or in Type1Font.cpp implements the "If an encoding maps to a character
name that does not exist in the Type 1 font pro- gram, the .notdef glyph
is substituted." line from the PDF 1.7 spec (in 5.5.5 Character
Encoding, Encodings for Type 1 Fonts) yet, so this does yet have an
effect.
This adapts our implementation to the editorial change in the temporal
proposal: https://github.com/tc39/proposal-temporal/commit/737baf2d
The changes to CalendarMethodsRecordLookup had already been implemented,
but we had followed the typo in the spec for CalendarMethodsRecordCall.
The issue in CalendarMethodsRecordCall hasn't surfaced yet, as the AOs
using Calendar Methods Record are currently not passing through a String
to represent a Calendar builtin.
No change to test-262.
When a node is removed from the DOM tree, its paintable needs to be
removed to ensure that it is not used to obtain sizes that are no
longer valid.
This change enables the ResizeObserver to send a notification if a node
is removed, as it should, because a removed node now has a size of zero
It should be okay to nullify pointers without concerning
parent/sibling/child relationships because the layout and paintable
trees will be rebuilt following any DOM mutation anyway.
Extends event loop processing steps to include gathering and
broadcasting resize observations.
Moves layout updates from Navigable::paint() to event loop processing
steps. This ensures resize observation processing occurs between layout
updates and painting.
In this change, updating layout and painting are moved to the EventLoop
processing steps. This modification allows the addition of resize
observation dispatching that needs to happen in event loop processing
steps and must occur in the following order relative to layout and
painting:
1. Update layout.
2. Gather and broadcast resize observations.
3. Paint.
Adds the initial implementation for interfaces defined in the
ResizeObserver specification. These interfaces will be used to
construct and send observation events in the upcoming changes.
This number is pure guesswork but it appears to fix GCC builds with
both ASAN and UBSAN hitting a native stack overflow before we have
a chance to catch it on our Azure CI.
We currently have a handful of iframe tests whose sources are in the
"input" directory. This means they get run as their own tests, when they
are really just helper files. We've had to add empty test expectation
files for these "tests", and invoke a dummy test() method just to keep
the test runner happy.
Instead, move them to their own directory so the test runner does not
see them at all.
This patch builds off the previous commit and removes the mouse hacks
entirely as they are not needed anymore.
With this, Sonic Robo Blast 2 now correctly locks the mouse inside the
game and makes for a much better playing experience! :^)
This patch implements the mouse warping functionality of SDL2.
This adds a WarpMouse function implementation to the SDL2 port.
SDL2 will then be able to use this for it's relative mouse mode
implementation.
With this, multiple ports depending on SDL2 now correctly lock the
mouse inside the window and it improves the experience significantly.
Note that as of now, you may need to pass the kernel argument
'vmmouse=off' in order to test these changes properly.
This implements support for `glBlendEquation` and
`glBlendEquationSeparate`. These functions modify the calculation of the
resulting color in blending mode.
This patch moves us away from the accumulator-based bytecode format to
one with explicit source and destination registers.
The new format has multiple benefits:
- ~25% faster on the Kraken and Octane benchmarks :^)
- Fewer instructions to accomplish the same thing
- Much easier for humans to read(!)
Because this change requires a fundamental shift in how bytecode is
generated, it is quite comprehensive.
Main implementation mechanism: generate_bytecode() virtual function now
takes an optional "preferred dst" operand, which allows callers to
communicate when they have an operand that would be optimal for the
result to go into. It also returns an optional "actual dst" operand,
which is where the completion value (if any) of the AST node is stored
after the node has "executed".
One thing of note that's new: because instructions can now take locals
as operands, this means we got rid of the GetLocal instruction.
A side-effect of that is we have to think about the temporal deadzone
(TDZ) a bit differently for locals (GetLocal would previously check
for empty values and interpret that as a TDZ access and throw).
We now insert special ThrowIfTDZ instructions in places where a local
access may be in the TDZ, to maintain the correct behavior.
There are a number of progressions and regressions from this test:
A number of async generator tests have been accidentally fixed while
converting the implementation to the new bytecode format. It didn't
seem useful to preserve bugs in the original code when converting it.
Some "does eval() return the correct completion value" tests have
regressed, in particular ones related to propagating the appropriate
completion after control flow statements like continue and break.
These are all fairly obscure issues, and I believe we can continue
working on them separately.
The net test262 result is a progression though. :^)
This is pure prep work for refactoring the bytecode to use more operands
instead of only registers.
generate_bytecode() virtuals now return an Optional<Operand>, and the
idea is to return an Operand referring to the value produced by this
AST node.
They also take an Optional<Operand> "preferred_dst" input. This is
intended to communicate the caller's preference for an output operand,
if any. This will be used to elide temporaries when we can store the
result directly in a local, for example.
The JIT compiler was an interesting experiment, but ultimately the
security & complexity cost of doing arbitrary code generation at runtime
is far too high.
In subsequent commits, the bytecode format will change drastically, and
instead of rewriting the JIT to fit the new bytecode, this patch simply
removes the JIT instead.
Other engines, JavaScriptCore in particular, have already proven that
it's possible to handle the vast majority of contemporary web content
with an interpreter. They are currently ~5x faster than us on benchmarks
when running without a JIT. We need to catch up to them before
considering performance techniques with a heavy security cost.
This can apply to optional, nullable parameters of platform types. This
will cause the generator to generate the type as
Optional<JS::GCPtr<T>> rather than JS::GCPtr<T>, allowing the
implementation to differentiate between the case where the argument was
not passed, and the case where null or undefined was passed.
Needing to differentiate these two cases is quite niche, hence why it is
an opt-in behavior.
