The spec has a note stating that resolve binding will always return a
reference whose [[ReferencedName]] field is name. However this is not
correct as the underlying method GetIdentifierReference may throw on
env.HasBinding(name) thus it can throw. However, there are some
scenarios where it cannot throw because the reference is known to exist
in that case we use MUST with a comment.
Exactly like in 99f9609, which fixed the same issue in CallExpression,
the spec tells us to *first* evaluate the arguments, if any, and *then*
check if the provided value is a constructor function.
Although those are the only valid options parse_primary_expression is
sometimes called when only an expression is valid which means it did not
check match_expression and might fail the now removed VERIFY.
Previously we might swallow invalid unicode point which would skip valid
ascii characters. This could be dangerous as we might skip a '"' thus
not closing a string where we should.
This might have been exploitable as it would not have been clear what
code gets executed when looking at a script.
Another approach to this would be simply replacing all invalid
characters with the replacement character (this is what v8 does). But
our lexer and parser are currently not set up for such a change.
The this value is only supposed to be set via the BindThisValue and
accessed via the GetThisBinding AOs, so exposing a direct getter/setter
would only lead to potentially non-spec-compliant behavior down the
line.
...for now - the reason being that the AST breaks 'completion bubbling'
and returns a plain Value, and code at the call site relies on the VM
having an exception set when converting the plain value back into a
completion.
Fixes#11301.
This implements:
- console.group()
- console.groupCollapsed()
- console.groupEnd()
In the Browser, we use `<details>` for the groups, which is not actually
implemented yet, so groups are always open.
In the REPL, groups are non-interactive, but still indent any output.
This looks weird since the console prompt and return values remain on
the far left, but this matches what Node does so it's probably fine. :^)
I expect `console.group()` is not used much outside of browsers.
The spec very kindly defines `Printer` as accepting
"Implementation-specific representations of printable things such as a
stack trace or group." for the `args`. We make use of that here by
passing the `Trace` itself to `Printer`, instead of having to produce a
representation of the stack trace in advance and then pass that to
`Printer`. That both avoids the hassle of tracking whether the data has
been html-encoded or not, and means clients don't have to implement the
whole `trace()` algorithm, but only the code needed to output the trace.
The `CountReset` log level is displayed as a warning, since the message
is always to warn that the counter doesn't exist. This is also in line
with the table at https://console.spec.whatwg.org/#loglevel-severity
This implements the Logger and Printer abstract operations defined in
the console spec, and stubs out the Formatter AO. These are then used
for the "output a categorized log message" functions.
The general idea when ENABLE_UNICODE_DATABASE_DOWNLOAD is OFF has been
that the Intl APIs will provide obviously incorrect results, but should
not crash. This regressed a bit with NumberFormat and DateTimeFormat.
Just like with integral and floating numbers, doing it manually is
error-prone: when we get a negative remainder, y has to be added to the
result to match the spec's modulo.
Solve this by just adding another (templated) overload to also permit
using the function for LibCrypto BigInts.
It's a bit annoying having to add '.0' to y given that it's an integral
number in most cases.
This turns the single template parameter T into T and U to permit that.
In order to reduce our reliance on __builtin_{ffs, clz, ctz, popcount},
this commit removes all calls to these functions and replaces them with
the equivalent functions in AK/BuiltinWrappers.h.
This also replaces an instance of TRY with MUST as the spec indicates
that step 35b of RegExpBuiltinExec cannot throw. Further, this moves
some lines of code around to align with the spec as best as we can,
though the end effect is the same.
In doing so, this caught an erroneous ToObject invocation. In the one
spot that requires the value to be an object (the invocation to
LengthOfArrayLike), we know by then the value is an object because all
other possibilities have been handled.
In doing so, this fixes a few minor issues:
1. We were accessing the "unicode" and "lastIndex" properties out of
order. This is somewhat frequently tested by test262, but not in
this case.
2. We were doing a Value::to_object() followed by Object::get(), rather
than just Value::get() as the spec dictates.
3. We were TRYing a step (CreateDataPropertyOrThrow) that should never
fail, and should have been a MUST.
RegExpExec already returns a ThrowCompletionOr so this potentional error
should be a completion. RegExp.prototype [ @@replace ] is tripped up by
this mistake when implemented closer to the spec.
