This note about how `v8::SnapshotCreator::create_blob` must not be
called from a `HandleScope` stopped being relevant in #6801, and was
now attached to code that had nothing to do with `HandleScope`s.
Streamlines a common middleware pattern and provides foundations for avoiding variably sized v8::ExternalReferences & enabling fully monomorphic op callpaths
When an exception is thrown during the processing of streaming WebAssembly,
`op_wasm_streaming_abort` is called. This op calls into V8, which synchronously
rejects the promise and calls into the promise rejection handler, if applicable.
But calling an op borrows the isolate's `JsRuntimeState` for the duration of the
op, which means it is borrowed when V8 calls into `promise_reject_callback`,
which tries to borrow it again, panicking.
This change changes `op_wasm_streaming_abort` from an op to a binding
(`Deno.core.abortWasmStreaming`). Although that binding must borrow the
`JsRuntimeState` in order to access the `WasmStreamingResource` stored in the
`OpTable`, it also takes ownership of that `WasmStreamingResource` instance,
which means it can drop any borrows of the `JsRuntimeState` before calling into
V8.
In the implementation of structured serialization in
`Deno.core.serialize`, whenever there is a serialization error, an
exception will be thrown with the message "Failed to serialize
response", even though V8 provides a message to use in such cases.
This change instead throws an exception with the V8-provided message,
if there is one.
This commit adds "--trace-ops" flag to "deno test" subcommand.
This flag enables saving of stack traces for async ops, that before were always
saved. While the feature proved to be very useful it comes with a significant performance
hit, it's caused by excessive source mapping of stack frames.
This commit improves the error messages for the `deno test` async op
sanitizer. It does this in two ways:
- it uses handwritten error messages for each op that could be leaking
- it includes traces showing where each op was started
This "async op tracing" functionality is a new feature in deno_core.
It likely has a significant performance impact, which is why it is only
enabled in tests.
This commit rewrites "InspectorSession" to no longer implement "Future"
trait but instead implement "Stream" trait. "Stream" trait is implemented
by yielding a raw pointer to the "v8::inspector::V8InspectorSession" and
received message. In effect received messages are no longer dispatched
from within the future, but are explicitly dispatched by the caller.
This change should allow us to dispatch a message to the session when
another message is being dispatched, ie.
"V8InspectorSesssion::dispatch_protocol_message" is already on the
call stack.
This commit changes flow in inspector code to no longer require
"Runtime.runIfWaitingForDebugger" message to complete a handshake.
Even though clients like Chrome DevTools always send this message on startup,
it is against the protocol to require this message to start an inspector
session.
Instead "Runtime.runIfWaitingForDebugger" is required only when running with
"--inspect-brk" flag, which matches behavior of Node.js.
`CrossIsolateStore`, `ExtensionBuilder` and `InMemoryChannelResource`
are private types which are referred to by other public APIs, and so
don't show up as links in the rustdoc. This is especially confusing for
`ExtensionBuilder`, since there is nothing in the docs that explains how
to build an extension.
Exposing these three types doesn't add any new capabilities:
`ExtensionBuilder` can be created from `Extension::builder()`,
`SharedArrayBufferStore` and `CompiledWasmModuleStore` already enable
doing anything that `CrossIsolateStore` can do by itself, and
`InMemoryChannelResource` isn't constructable.
This commit adds proper support for import assertions and JSON modules.
Implementation of "core/modules.rs" was changed to account for multiple possible
module types, instead of always assuming that the code is an "ES module". In
effect "ModuleMap" now has knowledge about each modules' type (stored via
"ModuleType" enum). Module loading pipeline now stores information about
expected module type for each request and validates that expected type matches
discovered module type based on file's "MediaType".
Relevant tests were added to "core/modules.rs" and integration tests,
additionally multiple WPT tests were enabled.
There are still some rough edges in the implementation and not all WPT were
enabled, due to:
a) unclear BOM handling in source code by "FileFetcher"
b) design limitation of Deno's "FileFetcher" that doesn't download the same
module multiple times in a single run
Co-authored-by: Kitson Kelly <me@kitsonkelly.com>
Provide a programmatic means of intercepting rejected promises without a
.catch() handler. Needed for Node compat mode.
Also do a first pass at uncaughtException support because they're
closely intertwined in Node. It's like that Frank Sinatra song:
you can't have one without the other.
Stepping stone for #7013.
This commit adds an ability to "ref" or "unref" pending ops.
Up to this point Deno had a notion of "async ops" and "unref async ops";
the former keep event loop alive, while the latter do not block event loop
from finishing. It was not possible to change between op types after
dispatching, one had to decide which type to use before dispatch.
Instead of storing ops in two separate "FuturesUnordered" collections,
now ops are stored in a single collection, with supplemental "HashSet"
storing ids of promises that were "unrefed".
Two APIs were added to "Deno.core":
"Deno.core.refOp(promiseId)" which allows to mark promise id
to be "refed" and keep event loop alive (the default behavior)
"Deno.core.unrefOp(promiseId)" which allows to mark promise
id as "unrefed" which won't block event loop from exiting
This commit adds several new "Deno.core" bindings:
* "setNextTickCallback"
* "hasScheduledTick"
* "setHasScheduledTick"
* "runMicrotasks"
Additionally it changes "Deno.core.setMacrotaskCallback" to
allow registering multiple callbacks. All these changes were necessary
to polyfill "process.nextTick" in Node compat layer.
Co-authored-by: Ben Noordhuis <info@bnoordhuis.nl>
This allows resources to be "streams" by implementing read/write/shutdown. These streams are implicit since their nature (read/write/duplex) isn't known until called, but we could easily add another method to explicitly tag resources as streams.
`op_read/op_write/op_shutdown` are now builtin ops provided by `deno_core`
Note: this current implementation is simple & straightforward but it results in an additional alloc per read/write call
Closes #12556
This adds `.code` attributes to errors returned by the op-layer, facilitating classifying OS errors and helping node-compat.
Similar to Node, these `.code` attributes are stringified names of unix ERRNOs, the mapping tables are generated by [tools/codegen_error_codes.js](https://gist.github.com/AaronO/dfa1106cc6c7e2a6ebe4dba9d5248858) and derived from libuv and rust's std internals
Currently all async ops are polled lazily, which means that op
initialization code is postponed until control is yielded to the event
loop. This has some weird consequences, e.g.
```js
let listener = Deno.listen(...);
let conn_promise = listener.accept();
listener.close();
// `BadResource` is thrown. A reasonable error would be `Interrupted`.
let conn = await conn_promise;
```
JavaScript promises are expected to be eagerly evaluated. This patch
makes ops actually do that.
WebAssembly modules compiled through `WebAssembly.compile()` and similar
non-streaming APIs don't have a URL associated to them, because they
have been compiled from a buffer source. In stack traces, V8 will use
a URL such as `wasm://wasm/d1c677ea`, with a hash of the module.
However, wasm modules compiled through streaming APIs, like
`WebAssembly.compileStreaming()`, do have a known URL, which can be
obtained from the `Response` object passed into the streaming APIs. And
as per the developer-facing display conventions in the WebAssembly
Web API spec, this URL should be used in stack traces. This change
implements that.
Decouple JsRuntime::sync_ops_cache() from the availability of the Deno.* namespace in the global scope
This avoids crashes when calling sync_ops_cache() on a bootstrapped WebWorker who has dropped its Deno.* namespace
It's also just cleaner and more robust ...
This commit fixes a problem where loading and executing multiple
modules leads to all of the having "import.meta.main" set to true.
Following Rust APIs were deprecated:
- deno_core::JsRuntime::load_module
- deno_runtime::Worker::execute_module
- deno_runtime::WebWorker::execute_module
Following Rust APIs were added:
- deno_core::JsRuntime::load_main_module
- deno_core::JsRuntime::load_side_module
- deno_runtime::Worker::execute_main_module
- deno_runtime::Worker::execute_side_module
- deno_runtime::WebWorker::execute_main_module
Trying to load multiple "main" modules into the runtime now results in an
error. If user needs to load additional "non-main" modules they should use
APIs for "side" module.
Async WebAssembly compilation was implemented by adding two
bindings: `set_wasm_streaming_callback`, which registered a callback to
be called whenever a streaming wasm compilation was started, and
`wasm_streaming_feed`, which let the JS callback modify the state of the
v8 wasm compiler.
`set_wasm_streaming_callback` cannot currently be implemented as
anything other than a binding, but `wasm_streaming_feed` does not really
need to use anything specific to bindings, and could indeed be
implemented as one or more ops. This PR does that, resulting in a
simplification of the relevant code.
There are three operations on the state of the v8 wasm compiler that
`wasm_streaming_feed` allowed: feeding new bytes into the compiler,
letting it know that there are no more bytes coming from the network,
and aborting the compilation. This PR provides `op_wasm_streaming_feed`
to feed new bytes into the compiler, and `op_wasm_streaming_abort` to
abort the compilation. It doesn't provide an op to let v8 know that the
response is finished, but closing the resource with `Deno.core.close()`
will achieve that.
Because it was possible to disable those with a runtime flag, they were
not available through primordials. The flag has since been removed
upstream.
Refs: d59db06bf5
* refactor(ops): return BadResource errors in ResourceTable calls
Instead of relying on callers to map Options to Results via `.ok_or_else(bad_resource_id)` at over 176 different call sites ...
Oneshot is more appropriate because mod_evaluate() only sends a single
value.
It also makes it easier to use it correctly. As an embedder, I wasn't
sure if I'm expected to drain the channel or not.
This commit changes return type of JsRuntime::execute_script to include
v8::Value returned from evaluation.
When embedding deno_core it is sometimes useful to be able to inspect
script evaluation value without the hoops of adding ops to store the
value on the OpState.
v8::Global<v8::Value> is used so consumers don't have to pass
scope themselves.
The WebAssembly streaming APIs used to be enabled, but used to take
buffer sources as their first argument (see #6154 and #7259). This
change re-enables them, requiring a Promise<Response> instead, as well as
enabling asynchronous compilation of WebAssembly modules.
This commit introduces primordials to deno_core. Primordials are a
frozen set of all intrinsic objects in the runtime. They are not
vulnerable to prototype pollution.
This commit changes implementation of module loading in "deno_core"
to track all currently fetched modules across all existing module loads.
In effect a bug that caused concurrent dynamic imports referencing the
same module to fail is fixed.
This commits moves implementation of net related APIs available on "Deno"
namespace to "deno_net" extension.
Following APIs were moved:
- Deno.listen()
- Deno.connect()
- Deno.listenTls()
- Deno.serveHttp()
- Deno.shutdown()
- Deno.resolveDns()
- Deno.listenDatagram()
- Deno.startTls()
- Deno.Conn
- Deno.Listener
- Deno.DatagramConn
This commit adds support for piping console messages to inspector.
This is done by "wrapping" Deno's console implementation with default
console provided by V8 by the means of "Deno.core.callConsole" binding.
Effectively each call to "console.*" methods calls a method on Deno's
console and V8's console.
Starting with V8 9.1, top-level-await is always enabled by default.
See https://v8.dev/blog/v8-release-91 for the release notes.
- Remove the now redundant v8 flag.
- Clarify doc comment and add link to the feature explainer.
This commit renames "JsRuntime::execute" to "JsRuntime::execute_script". Additionally
same renames were applied to methods on "deno_runtime::Worker" and
"deno_runtime::WebWorker".
A new macro was added to "deno_core" called "located_script_name" which
returns the name of Rust file alongside line no and col no of that call site.
This macro is useful in combination with "JsRuntime::execute_script"
and allows to provide accurate place where "one-off" JavaScript scripts
are executed for internal runtime functions.
Co-authored-by: Nayeem Rahman <nayeemrmn99@gmail.com>
This commit changes module loading implementation in "deno_core"
to call "ModuleLoader::prepare" hook only once per entry point.
This is done to avoid multiple type checking of the same code
in case of duplicated dynamic imports.
Relevant code in "cli/module_graph.rs" was updated as well.
This commit removes all JS based text encoding / text decoding. Instead
encoding now happens in Rust via encoding_rs (already in tree). This
implementation retains stream support, but adds the last missing
encodings. We are incredibly close to 100% WPT on text encoding now.
This should reduce our baseline heap by quite a bit.
This speeds up incremental rebuild when only touching JS files by 13-15%
Rebuild time after `touch 01_broadcast_channel.js`:
main: run 1 49.18s, run 2 50.34s
this: run 1 43.12s, run 2 43.19s
This commit moves implementation of "JsRuntimeInspector" to "deno_core" crate.
To achieve that following changes were made:
* "Worker" and "WebWorker" no longer own instance of "JsRuntimeInspector",
instead it is now owned by "deno_core::JsRuntime".
* Consequently polling of inspector is no longer done in "Worker"/"WebWorker",
instead it's done in "deno_core::JsRuntime::poll_event_loop".
* "deno_core::JsRuntime::poll_event_loop" and "deno_core::JsRuntime::run_event_loop",
now accept "wait_for_inspector" boolean that tells if event loop should still be
"pending" if there are active inspector sessions - this change fixes the problem
that inspector disconnects from the frontend and process exits once the code has
stopped executing.
This commit moves bulk of the logic related to module loading
from "JsRuntime" to "ModuleMap".
Next steps are to rewrite the actual loading logic (represented by
"RecursiveModuleLoad") to be a part of "ModuleMap" as well --
that way we will be able to track multiple module loads from within
the map which should help me solve the problem of concurrent
loads (since all info about currently loading/loaded modules will
be contained in the ModuleMap, so we'll be able to know if actually
all required modules have been loaded).
This ensures that provided extensions are all correctly setup and ready to use once the JsRuntime constructor returns
Note: this will also initialize ops for to-be-snapshotted runtimes
Extensions allow declarative extensions to "JsRuntime" (ops, state, JS or middleware).
This allows for:
- `op_crates` to be plug-and-play & self-contained, reducing complexity leaked to consumers
- op middleware (like metrics_op) to be opt-in and for new middleware (unstable, tracing,...)
- `MainWorker` and `WebWorker` to be composable, allowing users to extend workers with their ops whilst benefiting from the other infrastructure (inspector, etc...)
In short extensions improve deno's modularity, reducing complexity and leaky abstractions for embedders and the internal codebase.
General cleanup of module loading code, tried to reduce indentation in various methods
on "JsRuntime" to improve readability.
Added "JsRuntime::handle_scope" helper function, which returns a "v8::HandleScope".
This was done to reduce a code pattern that happens all over the "deno_core".
Additionally if event loop hangs during loading of dynamic modules a list of
currently pending dynamic imports is printed.
`InvalidDNSNameError` is thrown when a string is not a valid hostname,
e.g. it contains invalid characters, or starts with a numeric digit. It
does not involve a (failed) DNS lookup.
Even if bootstrapping the JS runtime is low level, it's an abstraction leak of
core to require users to call `Deno.core.ops()` in JS space.
So instead we're introducing a `JsRuntime::sync_ops_cache()` method,
once we have runtime extensions a new runtime will ensure the ops
cache is setup (for the provided extensions) and then loading/unloading
plugins should be the only operations that require op cache syncs
- register builtin v8 errors in core.js so consumers don't have to
- remove complexity of error args handling (consumers must provide a
constructor with custom args, core simply provides msg arg)
`init()` was previously needed to init the shared queue, but now that it's
gone `init()` only registers the async msg handler which is snapshot
safe and constant since the op layer refactor.
Also cleanup `bindings::deserialize()/decode()` so they
use the `ZeroCopyBuf` abstraction rather than reimplementing it.
This cleanup will facilitate moving `ZeroCopyBuf` to `serde_v8`
since it's now self contained and there are no other
`get_backing_store_slice()` callers.
This commit replaces GothamState's internal HashMap
with a BTreeMap to improve performance.
OpState/GothamState keys by TypeId which is essentially
an opaque u64. For small sets of integer keys BTreeMap
outperforms HashMap mainly since it removes the hashing
overhead and Eq/Comp on integer-like types is very cheap,
it should also have a smaller memory footprint.
We only use ~30 unique types and thus ~30 unique keys to
access OpState, so the keyset is small and immutable
throughout the life of a JsRuntime, there's no meaningful
churn in keys added/removed.
This is another optimization to help improve the baseline overhead
of async ops. It shaves off ~55ns/op or ~7% of the current total
async op overhead.
It achieves these gains by taking advantage of the sequential
nature of promise IDs and optimistically stores them sequentially
in a pre-allocated circular buffer and fallbacks to the promise Map
for slow to resolve promises.
- Improves op performance.
- Handle op-metadata (errors, promise IDs) explicitly in the op-layer vs
per op-encoding (aka: out-of-payload).
- Remove shared queue & custom "asyncHandlers", all async values are
returned in batches via js_recv_cb.
- The op-layer should be thought of as simple function calls with little
indirection or translation besides the conceptually straightforward
serde_v8 bijections.
- Preserve concepts of json/bin/min as semantic groups of their
inputs/outputs instead of their op-encoding strategy, preserving these
groups will also facilitate partial transitions over to v8 Fast API for the
"min" and "bin" groups
