Avoid expensive cross-hierarchy dynamic_cast from JS::Object to
UniversalGlobalScopeMixin on every microtask checkpoint.
Since UniversalGlobalScopeMixin is not in the JS::Object
inheritance chain, as<UniversalGlobalScopeMixin>(JS::Object&)
falls through to dynamic_cast, which is very costly. Profiling
showed this taking ~14% of total CPU time.
Add EnvironmentSettingsObject::universal_global_scope() backed
by a pointer cached eagerly during initialization.
entry_realm() was using the topmost execution context, but the spec
defines the entry execution context as the most recently pushed *realm*
execution context — the one owned by the environment settings object.
In a synchronous cross-window call, JS function calls push additional
execution contexts above the entry realm, causing the wrong realm to
be returned. Fix this by walking the stack to find the context that
matches its environment settings object's realm execution context.
Previously, destroyed-document tasks were forced to be runnable to
prevent them from leaking in the task queue. Instead, discard them
during task selection so their callbacks never run with stale state.
This used to cause issues with a couple of `spin_until()`s in the past,
but since we've removed some of them that had to do with the document
lifecycle, let's see if we can stick closer to the spec now.
Replace the BytecodeFactory header with cbindgen.
This will help ensure that types and enums and constants are kept in
sync between the C++ and Rust code. It's also a step in exporting more
Rust enums directly rather than relying on magic constants for
switch statements.
The FFI functions are now all placed in the JS::FFI namespace, which
is the cause for all the churn in the scripting parts of LibJS and
LibWeb.
Replace the 16-byte Variant<Empty, GC::Ref<Script>, GC::Ref<Module>>
with a simple 8-byte GC::Ptr<Cell> that points to either a Script or
Module (or is null for Empty).
A helper function script_or_module_from_cell() converts back to the
full ScriptOrModule variant when needed (e.g. in
VM::get_active_script_or_module).
This field was only used by LibWeb to prevent GC collection of the
EnvironmentSettingsObject while its execution context is on the stack.
This is unnecessary because the ESO is already reachable through the
realm's host_defined pointer: EC -> realm -> host_defined ->
PrincipalHostDefined -> environment_settings_object.
Shrinks ExecutionContext from 152 to 144 bytes.
When parse_off_thread() completes, the result callback runs inside a
deferred_invoke, which executes outside the HTML event loop's task
model. This meant that any microtasks queued by the callback (e.g.
promise reactions from react_to_promise during module linking) were
never drained, since HTML::EventLoop::process() only performs
microtask checkpoints after executing an HTML task.
Fix this by performing an explicit microtask checkpoint after the
parse result callback. This ensures that promise reactions queued
during module linking are drained immediately.
This fixes module worker scripts timing out because their loading
promise chains would stall indefinitely.
Use the parse_off_thread() helper to submit
parse_program(ProgramType::Module) to the ThreadPool for parsing
on a worker thread. Bounce back to the main thread to compile and
deliver the result via deferred_invoke.
Falls back to synchronous parsing when the Rust pipeline is
unavailable (LIBJS_CPP=1 or LIBJS_COMPARE_PIPELINES=1).
Add compile_parsed_module() to RustIntegration, which takes a
RustParsedProgram and a SourceCode (from parse_program with
ProgramType::Module) and compiles it on the main thread with GC
interaction.
Rewrite compile_module() to use the new split functions internally.
Add SourceTextModule::parse_from_pre_parsed() and
JavaScriptModuleScript::create_from_pre_parsed() to allow creating
module scripts from a pre-parsed RustParsedProgram.
This prepares the infrastructure for off-thread module parsing.
Create a SourceCode on the main thread (performing UTF-8 to UTF-16
conversion), then submit parse_program() to the ThreadPool for
Rust parsing on a worker thread. This unblocks the WebContent event
loop during external script loading.
Add Script::create_from_parsed() and
ClassicScript::create_from_pre_parsed() factory methods that take a
pre-parsed RustParsedProgram and a SourceCode, performing only the
GC-allocating compile step on the main thread.
Falls back to synchronous parsing when the Rust pipeline is
unavailable (LIBJS_CPP=1 or LIBJS_COMPARE_PIPELINES=1).
Replace alloca-based execution context allocation with InterpreterStack
bump allocation across all call sites: bytecode call instructions,
AbstractOperations call/construct, script evaluation, module evaluation,
and LibWeb module script evaluation.
Also replace the native stack space check with an InterpreterStack
exhaustion check, and remove the now-unused alloca macros from
ExecutionContext.h.
Fixes the included imported test. Note that this required a minor
edit of the WPT import to work with our test harness setup to
try and create a non secure context setup as both file:// and
localhost are considered secure contexts.
This is somewhat awkward as the spec refers to 'is secure context'
with respect to these objects 'relevant settings object'. A natural
way of implementing this could be storing a pointer to the relevant
settings object like the JS representations of these objects do
(and then changing is_secure_context to accept this representation
too), but for now it seems much simpler to just store a boolean for
this purpose and sidestep both problems above.
Instead of passing through window's associated document's URL as
an extra argument to starting up a worker. This will allow for
improving the representation of 'outside settings' when setting
up a Worker.
The m_result Variant can hold a GC::Ref<ImportMapParseResult> when the
script element has type="importmap", but visit_edges only traced the
GC::Ref<Script> arm. This left the ImportMapParseResult unvisited,
allowing the GC to collect it while the element still held a reference.
ImportMapParseResult inherited from JS::Script::HostDefined, but no
JS::Script or JS::Module ever stored it as host_defined data, so
visit_host_defined_self was dead code. This removes the HostDefined
inheritance entirely and switches m_result visitation to Variant::visit
with a lambda that catches all GC::Ref arms.
As a stopgap until we have things wired up fully, consider
WorkerGlobalScope as a SecureContext so that IDL marked with
[SecureContext] is still exposed.
Change the parameters types of the functions `coarsen_time` and
`coarsened_shared_current_time` from `bool` to
`CanUseCrossOriginIsolatedAPIs` for more coherence with the surrounding
code.
The full constructor for NotificationsAPI::Notification is implemented
along with the getter methods.
It is now possible to call the elements of Notification in JS without
getting undefined but the default values (or the ones passed in
options).
The method `current_wall_time` is added in EnvironmentSettingsObject.
This passes a least a few more tests because of the getter methods
that are created.
https://wpt.live/notifications/constructor-basic.https.htmlhttps://wpt.live/notifications/idlharness.https.any.html
Remove 11 heavy includes from Document.h that were only needed for
pointer/reference types (already forward-declared in Forward.h), and
extract the nested ViewportClient interface to a standalone header.
This reduces Document.h's recompilation cascade from ~1228 files to
~717 files (42% reduction). Headers like BrowsingContext.h that were
previously transitively included see even larger improvements (from
~1228 down to ~73 dependents).
After removing the unwind context stack, ExecutionContextRareData only
held two GC::Ptr fields — both trivially destructible. The indirection
cost more than it saved: a GC cell allocation per EC, an extra pointer
chase on every source range lookup, and unnecessary complexity.
Replace the rare data cell with two inline fields on ExecutionContext:
cached_source_range and context_owner.
...and WorkerEnvironmentSettingsObject.h
These headers only use Document via forward-declarable references and
smart pointers, so the full include is unnecessary.
This is expected by WPT. For this to work, we must be able to determine
the network partition key for shared worker environments. So we now set
a top-level origin for these environments, with a FIXME to implement it
in accordance with the Client-Side Storage Partitioning spec.
This ensures that we are explicitly declaring the allocator to use when
allocating a cell(-inheriting) type, instead of silently falling back
to size-based allocation.
Since this is done in allocate_cell, this will only be detected for
types that are actively being allocated. However, since that means
they're _not_ being allocated, that means it's safe to not declare
an allocator to use for those. For example, the base TypedArray<T>,
which is never directly allocated and only the defined specializations
are ever allocated.
Every function call allocates an ExecutionContext with a trailing array
of Values for registers, locals, constants, and arguments. Previously,
the constructor would initialize all slots to js_special_empty_value(),
but constant slots were then immediately overwritten by the interpreter
copying in values from the Executable before execution began.
To eliminate this redundant initialization, we rearrange the layout from
[registers | constants | locals] to [registers | locals | constants].
This groups registers and locals together at the front, allowing us to
initialize only those slots while leaving constant slots uninitialized
until they're populated with their actual values.
This reduces the per-call initialization cost from O(registers + locals
+ constants) to O(registers + locals).
Also tightens up the types involved (size_t -> u32) and adds VERIFYs to
guard against overflow when computing the combined slot counts, and to
ensure the total fits within the 29-bit operand index field.
Issue #6294 describes an edge case where the browser crash if the same
module is loaded three times in a document, but all attempts fail.
Failure scenario:
1. Module load 1 set the state to "Fetching"
2. Module load 2 registers a callback to `on_complete` since the
current state is "Fetching"
3. Module load 1 finish with a failure, invoking the callback for load
number 2
4. Module load 3 cause a crash. The state is neither "Fetching" or
"ModuleScript", so we'll reset the state to "Fetching". This invokes
the callback for module load 2 again, now with an unexpected state
which will cause an assert violation.
Proposed fix is to remove the condition that invokes `on_complete`
immediately for successfully loaded modules only, the callback should
be invoked regardless of whether the fetch succeeded or failed.
This reveals a separate bug in HTMLScriptElement, where
`mark_as_ready()` can be invoked before
`m_steps_to_run_when_the_result_is_ready` is assigned.
This appears to be a spec bug, reported as
https://github.com/whatwg/html/issues/12073 and addressed by delaying
the callback by a task, similar to the issue was resolved for inline
scripts.
The end goal here is for LibHTTP to be the home of our RFC 9111 (HTTP
caching) implementation. We currently have one implementation in LibWeb
for our in-memory cache and another in RequestServer for our disk cache.
The implementations both largely revolve around interacting with HTTP
headers. But in LibWeb, we are using Fetch's header infra, and in RS we
are using are home-grown header infra from LibHTTP.
So to give these a common denominator, this patch replaces the LibHTTP
implementation with Fetch's infra. Our existing LibHTTP implementation
was not particularly compliant with any spec, so this at least gives us
a standards-based common implementation.
This migration also required moving a handful of other Fetch AOs over
to LibHTTP. (It turns out these AOs were all from the Fetch/Infra/HTTP
folder, so perhaps it makes sense for LibHTTP to be the implementation
of that entire set of facilities.)
An upcoming commit will migrate the contents of Headers.h/cpp to LibHTTP
for use outside of LibWeb. These CORS and MIME helpers depend on other
LibWeb facilities, however, so they cannot be moved.
We need to prevent these mutation observers from being garbage
collected, and since they are only part of SimilarOriginWindowAgent and
themselves as part of the intrusive list, nobody is visiting them.
Make the list of pending mutation observers a GC::RootVector so we keep
them alive until they have been processed in the microtask.
Restores 1400+ WPT subtest passes in `dom/nodes/Element-classlist.html`.
In the current spec, MutationObservers are explicitly added to the
pending mutation observers list, and they are removed when that list is
cleared in the "notify mutation observers" microtask.
This solves some issues with slotchange events.
As noted, we delay actually emptying the list of pending mutation
observers until after we're finished with the "clone", because we can't
actually copy or move the intrusive list. As far as I am aware, this
should not affect behaviour because only one microtask can run at once.
187f8c54 made `HTML::Task` runnable for destroyed documents, and this
change aligns microtask behavior with that. This is required for an
upcoming change that switches Fetch to be unbuffered by default. During
navigation, fetching the new document is initiated by the previous
document, which means we need to allow microtasks created in the
previous document's realm to run even after that document has been
destroyed.
We have a couple of ways to designate spec notes and (our) developer
notes in comments, but we never really settled on a single approach. As
a result, we have a bit of a mixed bag of note comments on our hands.
To the extent that I could find them, I changed developer notes to
`// NB: ...` and changed spec notes to `// NOTE: ...`. The rationale for
this is that in most web specs, notes are prefixed by `NOTE: ...` so
this makes it easier to copy paste verbatim. The choice for `NB: ...` is
pretty arbitrary, but it makes it stand out from the regular spec notes
and it was already in wide use in our codebase.
Atlassian login gets the base URL for its module scripts by throwing an
error and pulling out the current script's URL from error.stack with
regex.
Since we only returned a basename for module scripts, it would fail to
match and try and use `/` as a base URL (because it does
[matched_string] + "/"), which is not a valid base URL.
