Calling `worker.terminate()` used to kill the worker's isolate and
then block until the worker's thread finished. This blocks the calling
thread if the worker's event loop was blocked in a sync op (as with
`Deno.sleepSync`), which wasn't realized at the time, but since the
worker's isolate was killed at that moment, it would not block the
calling thread if the worker was in a JS endless loop.
However, in #12831, in order to work around a V8 bug, worker
termination was changed to first set a signal to let the worker event
loop know that termination has been requested, and only kill the
isolate if the event loop has not finished after 2 seconds. However,
this change kept the blocking, which meant that JS endless loops in
the worker now blocked the parent for 2 seconds.
As it turns out, after #12831 it is fine to signal termination and
even kill the worker's isolate without waiting for the thread to
finish, so this change does that. However, that might leave the async
ops that receive messages and control data from the worker pending
after `worker.terminate()`, which leads to odd results from the op
sanitizer. Therefore, we set up a `CancelHandler` to cancel those ops
when the worker is terminated.
This commit:
- removes "fmt_errors::PrettyJsError" in favor of "format_js_error" fn
- removes "deno_core::JsError::create" and
"deno_core::RuntimeOptions::js_error_create_fn"
- adds new option to "deno_runtime::ops::worker_host::init"
This commit changes "deno bench" subcommand, by updating
the "Deno.bench" API as follows:
- remove "Deno.BenchDefinition.n"
- remove "Deno.BenchDefintion.warmup"
- add "Deno.BenchDefinition.group"
- add "Deno.BenchDefintion.baseline"
This is done because bench cases are no longer run fixed amount
of iterations, but instead they are run until there is difference between
subsequent runs that is statistically insiginificant.
Additionally, console reporter was rewritten completely, to looks
similar to "hyperfine" reporter.
This commit rewrites test runner to send structured error data from JavaScript
to Rust instead of passing strings. This will allow to customize display of errors
in test report (which will be addressed in follow up commits).
The following transformations gradually faced by "JsError" have all been
moved up front to "JsError::from_v8_exception()":
- finding the first non-"deno:" source line;
- moving "JsError::script_resource_name" etc. into the first error stack
in case of syntax errors;
- source mapping "JsError::script_resource_name" etc. when wrapping
the error even though the frame locations are source mapped earlier;
- removing "JsError::{script_resource_name,line_number,start_column,end_column}"
entirely in favour of "js_error.frames.get(0)".
We also no longer pass a js-side callback to "core/02_error.js" from cli.
I avoided doing this on previous occasions because the source map lookups
were in an awkward place.
This commit changes "deno test" to better denote user output coming
from test cases.
This is done by printing "---- output ----" and "---- output end ----"
markers if an output is produced. The output from "console" and
"Deno.core.print" is captured, as well as direct writes to "Deno.stdout"
and "Deno.stderr".
To achieve that new APIs were added to "deno_core" crate, that allow
to replace an existing resource with a different one (while keeping resource
ids intact). Resources for stdout and stderr are replaced by pipes.
Co-authored-by: David Sherret <dsherret@gmail.com>
This commit adds following fields to "Deno.TestContext" interface:
- name
- origin
- parent
These are prerequisites for supporting snapshot functionality in
"std/testing".
Streamlines a common middleware pattern and provides foundations for avoiding variably sized v8::ExternalReferences & enabling fully monomorphic op callpaths
This commit adds "Deno.upgradeHttp" API, which
allows to "hijack" connection and switch protocols, to eg.
implement WebSocket required for Node compat.
Co-authored-by: crowlkats <crowlkats@toaxl.com>
Co-authored-by: Ryan Dahl <ry@tinyclouds.org>
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
Previously specifying permissions: {} was the same as specifying
permissions: "inherit". Now it will be the same as permissions: "none".
Not specifying any permissions (permissions: undefined) still means
permissions: "inherit".
This commit adds "deno bench" subcommand and "Deno.bench()"
API that allows to register bench cases.
The API is modelled after "Deno.test()" and "deno test" subcommand.
Currently the output is rudimentary and bench cases and not
subject to "ops" and "resource" sanitizers.
Co-authored-by: evan <github@evan.lol>
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 adds `readable` and `writable` properties to `Deno.File` and
`Deno.Conn`. This makes it very simple to use files and network sockets
with fetch or the native HTTP server.
Adds another callback to WebWorkerOptions that allows to execute
some modules before actual worker code executes. This allows to set up Node
global using std/node.
This commit makes the errors produced from the resource sanitizer much
more human readable. It does this by using real words rather than our
"resource names" when referring to resources, and by giving helpful
hints on how to clean up each of the resources.
This commit fixes an error when user deletes "window" global JS
variable. Instead of relying on "window" or "globalThis" to dispatch
"load" and "unload" events, we are default to global scope of the
worker.
Add an op to list the network interfaces on the system.
Prep work for #8137 and `os.networkInterfaces()` Node compat in std.
Refs denoland/deno_std#1436.
This commit fixes prompts printed to the terminal when
running with "--inspect" or "--inspect-brk" flags.
When debugger disconnects error is no longer printed as
users don't care about the reason debugger did disconnect.
A message suggesting to go to "chrome://inspect" is printed
if debugger is active.
Additionally and information that process is waiting for
debugger to connect is printed if running with "--inspect-brk"
flag.
This commit fixes inspector integration with "deno test" subcommand
by waiting for inspector sessions to connect if "--inspect-brk" flag
is passed.
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
Fixes "op_set_exit_code" by sharing a single "Arc" between
all workers (via "op state") instead of having a "global" value stored in
"deno_runtime" crate. As a consequence setting an exit code is always
scoped to a tree of workers, instead of being overridable if there are
multiple worker tree (like in "deno test --jobs" subcommand).
Refactored "cli/main.rs" functions to return "Result<i32, AnyError>" instead
of "Result<(), AnyError>" so they can return exit code.
Although not easy to replicate in the wild, the `deno test` op sanitizer
can fail when there are intervals that started before a test runs, since
the op sanitizer can end up running in the time between the timer op for
an interval's run resolves and the op for the next run starts.
This change fixes that by adding a new macrotask callback that will run
after the timer macrotask queue has drained. This ensures that there is
a timer op if there are any timers which are unresolved by the time the
op sanitizer runs.
Due to a bug in V8, terminating an isolate while a module with top-level
await is being evaluated would crash the process. This change makes it
so calling `worker.terminate()` will signal the worker to terminate at
the next iteration of the event loop, and it schedules a proper
termination of the worker's isolate after 2 seconds.
deno_fetch::init has a lot of parameters and generic on two types
that keeps expanding over time. This refactor adds deno_fetch::Options
struct for more clearly defining the various parameters.
Although not easy to replicate in the wild, the `deno test` op sanitizer
can fail when there are intervals that started before a test runs, since
the op sanitizer can end up running in the time between the timer op for
an interval's run resolves and the op for the next run starts.
This change fixes that by adding a new macrotask callback that will run
after the timer macrotask queue has drained. This ensures that there is
a timer op if there are any timers which are unresolved by the time the
op sanitizer runs.
Set the exit code to use if none is provided to Deno.exit(), or when
Deno exits naturally.
Needed for process.exitCode Node compat. Paves the way for #12888.
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
In tests, the function to format errors would assume that any error with
a property `errors` would be an `AggregateError`, and therefore the
property `errors` would contain an error. This is not necessarily the
case.
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
The initial implementation of `importScripts()` in #11338 used
`reqwest`'s default client to fetch HTTP scripts, which meant it would
not use certificates or other fetching configuration passed by command
line flags. This change fixes it.
A bug was fixed that could cause a hang when a method was
called on a TlsConn object that had thrown an exception earlier.
Additionally, a bug was fixed that caused TlsConn.write() to not
completely flush large buffers (>64kB) to the socket.
The public `TlsConn.handshake()` API is scheduled for inclusion in the
next minor release. See https://github.com/denoland/deno/pull/12467.
`Window`'s `self` property and `DedicatedWorkerGlobalScope`'s `name`
property are defined as Web IDL read-only attributes with the
`[Replaceable]` extended attribute, meaning that their setter will
redefine the property as a data property with the set value, rather than
changing some internal state. Deno currently defines them as read-only
data properties instead.
Given that Web IDL requires all attributes to be accessor properties
rather than data properties, but Deno exposes almost all of those
properties as either read-only or writable data properties, it makes
sense to expose `[Replaceable]` properties as writable as well – as is
already the case with `WindowOrWorkerGlobalScope`'s `performance`
property.
This commit annotates errors returned from FS Deno APIs to include
paths that were passed to the API calls.
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
These are confusing. They say they are "for users that don't care about
permissions", but that isn't correct. `NoTimersPermissions` disables
permissions instead of enabling them.
I would argue that implementors should decide what permissions they want
themselves, and not take our opinionated permissions struct.
* perf(ops): optimize permission check
Removes the overhead of permission check on access granted (should be common case):
Delta measured on `perf_now` from `deno_common` bench:
- before: `528ns/op
- after: `166ns/op`
So ~3x faster
This panic could happen in the following cases:
- A non-fatal error being thrown from a worker, that doesn't terminate
the worker's execution, but propagates to the main thread without
being handled, and makes the main thread terminate.
- A nested worker being alive while its parent worker gets terminated.
- A race condition if the main event loop terminates the worker as part
of its last task, but the worker doesn't fully terminate before the
main event loop stops running.
This panic happens because a worker's event loop should have pending ops
as long as the worker isn't closed or terminated – but if an event loop
finishes running while it has living workers, its associated
`WorkerThread` structs will be dropped, closing the channels that keep
those ops pending.
This change adds a `Drop` implementation to `WorkerThread`, which
terminates the worker without waiting for a response. This fixes the
panic, and makes it so nested workers are automatically terminated once
any of their ancestors is closed or terminated.
This change also refactors a worker's termination code into a
`WorkerThread::terminate()` method.
Closes #11342.
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
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.
* perf(runtime/fs): optimize readFile by using a single large buffer
* handle extended/truncated files during read
Allocate an extra byte in our read buffer to detect "overflow" then fallback to unsized readAll for remainder of extended file, this is a slowpath that should rarely happen in practice
A comment in `runtime.js` reads that `console` seems to be "the only one
that should be writable and non-enumerable", which explains why it is
declared with `util.writable` but then has its property descriptor's
`enumerable` key changed to false.
But it is not in fact true that `console` is the only global property
for which this holds, and it wasn't even when this behavior was
introduced in denoland#9013. All WebIDL interfaces are also writable and
non-enumerable – the only difference here being that `console` is a
namespace rather than an interface.
Since WebIDL interfaces are defined with `util.nonEnumerable`, and
`console` uses the same descriptor keys, this PR changes the definition
of `console` to use `util.nonEnumerable` as well.
This adds support for the URLPattern API.
The API is added in --unstable only, as it has not yet shipped in any
browser. It is targeted for shipping in Chrome 95.
Spec: https://wicg.github.io/urlpattern/
Co-authored-by: crowlKats < crowlkats@toaxl.com >
When `worker.terminate()` is called, the spec requires that the
corresponding port message queue is emptied, so no messages can be
received after the call, even if they were sent from the worker before
it was terminated.
The spec doesn't require this of `self.close()`, and since Deno uses
different channels to send messages and to notify that the worker was
closed, messages might still arrive after the worker is known to be
closed, which are currently being dropped. This change fixes that.
The fix involves two parts: one on the JS side and one on the Rust side.
The JS side was using the `#terminated` flag to keep track of whether
the worker is known to be closed, without distinguishing whether further
messages should be dropped or not. This PR changes that flag to an
enum `#state`, which can be one of `"RUNNING"`, `"CLOSED"` or
`"TERMINATED"`.
The Rust side was removing the `WorkerThread` struct from the workers
table when a close control was received, regardless of whether there
were any messages left to read, which made any subsequent calls to
`op_host_recv_message` to return `Ok(None)`, as if there were no more
mesasges. This change instead waits for both a close control and for
the message channel's sender to be closed before the worker thread is
removed from the table.
Classic worker scripts are now executed in the context of a Tokio
runtime. This does mean we can not spawn more tokio runtimes in
"op_worker_sync_fetch". We instead spawn a new thread there, that can
create a new Tokio runtime that we can use to block the worker thread.
This commit implements classic workers, but only when the `--enable-testing-features-do-not-use` flag is provided. This change is not user facing. Classic workers are used extensively in WPT tests. The classic workers do not support loading from disk, and do not support TypeScript.
Co-authored-by: Luca Casonato <hello@lcas.dev>
* 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 ...
This commit adds "--unsafely-treat-insecure-origin-as-secure" flag
that allows to disable SSL verification for all domains, or specific
domains if they were passed as an argument to the flag.
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
This commit adds "DENO_TLS_CA_STORE" env variable to support
optionally loading certificates from the users local certificate store.
This will allow them to successfully connect via tls with corporate
and self signed certs provided they have them installed in their keystore.
It also allows them to deal with revoked certs by simply updating
their keystore without having to upgrade Deno.
Currently supported values are "mozilla", "system" or empty value.
This commit removes implementation of "native plugins" and replaces
it with FFI API.
Effectively "Deno.openPlugin" API was replaced with "Deno.dlopen" API.
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 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.
This commit adds new options to unstable "Deno.createHttpClient" API.
"proxy" and "basicAuth" options were added that allow to use custom proxy
when client instance is passed to "fetch" API.
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 "op_http_response_write" to first send response chunk
and then poll the underlying HTTP connection.
Previously after writing a chunk of response HTTP connection wasn't polled
and thus data wasn't written to the socket until after next op interacting
with the connection.
Waiting on next request in Deno.serveHttp() API hanged
when responses were using ReadableStream. This was caused
by op_http_request_next op that was never woken after
response was fully written. This commit adds waker field to
DenoService which is called after response is finished.
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 commit adds "CancelHandle" to "ConnResource" and changes
"op_http_next_request" to await for the cancel signal. In turn
when async iterating over "Deno.HttpConn" the iterator breaks
upon closing of the resource.
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 refactors implementation of inspector.
The intention is to be able to move inspector implementation to "deno_core".
Following things were done to make that possible:
* "runtime/inspector.rs" was split into "runtime/inspector/mod.rs"
and "runtime/inspector/server.rs", separating inspector implementation
from Websocket server implementation.
* "DenoInspector" was renamed to "JsRuntimeInspector" and reference to "server"
was removed from the structure, making it independent of Websocket server
used to connect to Chrome Devtools.
* "WebsocketSession" was renamed to "InspectorSession" and rewritten in such
a way that it's not tied to Websockets anymore; instead it accepts a pair
of "proxy" channel ends that allow to integrate the session with different
"transports".
* "InspectorSession" was renamed to "LocalInspectorSession" to better indicate
that it's an "in-memory" session and doesn't require Websocket server. It was
also rewritten in such a way that it uses "InspectorSession" from previous point
instead of reimplementing "v8::inspector::ChannelImpl" trait; this is done by using
the "proxy" channels to communicate with the V8 session.
Consequently "LocalInspectorSession" is now a frontend to "InspectorSession". This
introduces a small inconvenience that awaiting responses for "LocalInspectorSession" requires
to concurrently poll worker's event loop. This arises from the fact that "InspectorSession"
is now owned by "JsRuntimeInspector", which in turn is owned by "Worker" or "WebWorker".
To ease this situation "Worker::with_event_loop" helper method was added, that takes
a future and concurrently polls it along with the event loop (using "tokio::select!" macro
inside a loop).
Replaces the file-backed provider by an in-memory one because proper
file locking is a hard problem that detracts from the proof of concept.
Teach the WPT runner how to extract tests from .html files because all
the relevant tests in test_util/wpt/webmessaging/broadcastchannel are
inside basics.html and interface.html.
In #9118, TLS streams were split into a "read half" and a "write half"
using tokio::io::split() to allow concurrent Conn#read() and
Conn#write() calls without one blocking the other. However, this
introduced a bug: outgoing data gets discarded when the TLS stream is
gracefully closed, because the read half is closed too early, before all
TLS control data has been received.
Fixes: #9692
Fixes: #10049
Fixes: #10296
Fixes: denoland/deno_std#750
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
This commit adds support for running test in parallel.
Entire test runner functionality has been rewritten
from JavaScript to Rust and a set of ops was added to support reporting in Rust.
A new "--jobs" flag was added to "deno test" that allows to configure
how many threads will be used. When given no value it defaults to 2.
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.
`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.
This commits adds adds "permissions" option to the test definitions
which allows tests to run with different permission sets than
the process's permission.
The change will only be in effect within the test function, once the
test has completed the original process permission set is restored.
Test permissions cannot exceed the process's permission.
You can only narrow or drop permissions, failure to acquire a
permission results in an error being thrown and the test case will fail.
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)
This commit aligns the `fetch` API and the `Request` / `Response`
classes belonging to it to the spec. This commit enables all the
relevant `fetch` WPT tests. Spec compliance is now at around 90%.
Performance is essentially identical now (within 1% of 1.9.0).
This commit fixes the URL returned from `request.url` in the HTTP server
to be fully qualified. This previously existed, but was removed and
accidentially not readded during optimizations of the HTTP ops.
Returning a non fully qualified URL from `Request#url` is not spec
compliant.
The panic was caused by the lack of an error class mapping for
futures::channel::TrySendError, but it shouldn't have been throwing an error in
the first place - when a worker has terminated, postMessage should just return.
The issue was that the termination message hadn't yet been recieved, so it was
carrying on with trying to send the message. This adds another check on the Rust
side for if the channel is closed, and if it is the worker is treated as
terminated.
