This commit completely overhauls how module analysis is
performed in TS compiler by moving the logic to Rust.
In the current setup module analysis is performed using
"ts.preProcessFile" API in a special TS compiler worker
running on a separate thread.
"ts.preProcessFile" allowed us to build a lot of functionality
in CLI including X-TypeScript-Types header support
and @deno-types directive support. Unfortunately at the
same time complexity of the ops required to perform
supporting tasks exploded and caused some hidden
permission escapes.
This PR introduces "ModuleGraphLoader" which can parse
source and load recursively all dependent source files; as
well as declaration files. All dependencies used in TS
compiler and now fetched and collected upfront in Rust
before spinning up TS compiler.
To achieve feature parity with existing APIs this commit
includes a lot of changes:
* add "ModuleGraphLoader"
- can fetch local and remote sources
- parses source code using SWC and extracts imports, exports, file references, special
headers
- this struct inherited all of the hidden complexity and cruft from TS version and requires
several follow up PRs
* rewrite cli/tsc.rs to perform module analysis upfront and send all required source code to
TS worker in one message
* remove op_resolve_modules and op_fetch_source_files from cli/ops/compiler.rs
* run TS worker on the same thread
This PR hot-fixes permission escapes in dynamic imports, workers
and runtime compiler APIs.
"permissions" parameter was added to public APIs of SourceFileFetcher
and appropriate permission checks are performed during loading of
local and remote files.
Importing .wasm files is non-standardized therefore deciding to
support current functionality past 1.0 release is risky.
Besides that .wasm import posed many challenges in our codebase
due to complex interactions with TS compiler which spawned
thread for each encountered .wasm import.
This commit removes:
- cli/compilers/wasm.rs
- cli/compilers/wasm_wrap.js
- two integration tests related to .wasm imports
This PR removes op_cache and refactors how Deno interacts with TS compiler.
Ultimate goal is to completely sandbox TS compiler worker; it should operate on
simple request -> response basis. With this commit TS compiler no longer
caches compiled sources as they are generated but rather collects all sources
and sends them back to Rust when compilation is done.
Additionally "Diagnostic" and its children got refactored to use "Deserialize" trait
instead of manually implementing JSON deserialization.
This change is to prevent needed a separate stat syscall for each file
when using readdir.
For consistency, this PR also modifies std's `WalkEntry` interface to
extend `DirEntry` with an additional `path` field.
* Properly track isFile, isSymlink, isDirectory
These don't exhaust all the possibilities, so none of them should be
defined as "neither of the others".
* empty
This is a first pass implementation which is still missing several important
features:
- support for --inspect-brk (#4503)
- support for source maps (#4501)
- support for piping console.log to devtools console (#4502)
Co-authored-by: Bert Belder <bertbelder@gmail.com>
Co-authored-by: Matt Harrison <mt.harrison86@gmail.com>
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
This a complex boring PR that shifts around code (primarily) in cli/fs.rs and
cli/ops/fs.rs. The gain of this refactoring is to ease the way for #4188 and
#4017, and also to avoid some future development pain.
Mostly there is no change in functionality. Except:
* squashed bugs where op_utime and op_chown weren't using `resolve_from_cwd`
* eliminated the use of the external `remove_dir_all` crate.
* op_chmod now only queries metadata to verify file/dir exists on Windows (it
will already fail on Unix if it doesn't)
* op_chown now verifies the file/dir's existence on Windows like chmod does.
Fixes #4101
Previously, we would just provide the raw JSON to the TypeScript
compiler worker, but TypeScript does not transform JSON. This caused
a problem when emitting a bundle, that the JSON would just be "inlined"
into the output, instead of being transformed into a module.
This fixes this problem by providing the compiled JSON to the TypeScript
compiler, so TypeScript just sees JSON as a "normal" TypeScript module.
Listener and UDPConn are AsyncIterables instead of AsyncIterators.
The [Symbol.asyncIterator]()s are defined as generators and the
next() methods are gone.
"Listener/Socket has been closed" errors are now BadResource.
There's a lot of variation in doc comments and internal code about
whether the first parameter to file system calls is `path` or `name` or
`filename`. For consistency, have made it always be `path`.
Rewrite "normalize_path()" to remove all intermediate components from the path, ie. "./" and "../". It's very similar in functionality to fs::canonicalize(), however "normalize_path() doesn't resolve symlinks.
To better reflect changes in error types in JS from #3662 this PR changes
default error type used in ops from "ErrBox" to "OpError".
"OpError" is a type that can be sent over to JSON; it has all
information needed to construct error in JavaScript. That
made "GetErrorKind" trait useless and so it was removed altogether.
To provide compatibility with previous use of "ErrBox" an implementation of
"From<ErrBox> for OpError" was added, however, it is an escape hatch and
ops implementors should strive to use "OpError" directly.
Example:
$ python2 -c 'open("\x80\x7F", "w")'
$ deno eval 'Deno.readDirSync(".")'
thread 'main' panicked at 'called `Option::unwrap()` on a `None` value', cli/ops/fs.rs:373:16
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
fatal runtime error: failed to initiate panic, error 5
Aborted (core dumped)
Before this commit they made deno panic, now they are silently skipped.
Not ideal but arguably better than panicking.
No test because what characters are and aren't allowed in filenames is
highly file system-dependent.
Closes #3950
* establish basic event loop for workers
* make "self.close()" inside worker
* remove "runWorkerMessageLoop() - instead manually call global function
in Rust when message arrives. This is done in preparation for structured clone
* refactor "WorkerChannel" and use distinct structs for internal
and external channels; "WorkerChannelsInternal" and "WorkerHandle"
* move "State.worker_channels_internal" to "Worker.internal_channels"
* add "WorkerEvent" enum for child->host communication;
currently "Message(Buf)" and "Error(ErrBox)" variants are supported
* add tests for nested workers
* add tests for worker throwing error on startup
This change simplifies how we execute V8. Previously V8 Isolates jumped
around threads every time they were woken up. This was overly complex and
potentially hurting performance in a myriad ways. Now isolates run on
their own dedicated thread and never move.
- blocking_json spawns a thread and does not use a thread pool
- op_host_poll_worker and op_host_resume_worker are non-operational
- removes Worker::get_message and Worker::post_message
- ThreadSafeState::workers table contains WorkerChannel entries instead
of actual Worker instances.
- MainWorker and CompilerWorker are no longer Futures.
- The multi-threaded version of deno_core_http_bench was removed.
- AyncOps no longer need to be Send + Sync
This PR is very large and several tests were disabled to speed
integration:
- installer_test_local_module_run
- installer_test_remote_module_run
- _015_duplicate_parallel_import
- _026_workers
For some reason, the unit tests for Deno.remove() were not being imported to
unit_tests.ts and, consequently, not being executed. Thus, I imported them,
refactored some existent ones and wrote new ones for the symlink removal case.
Since the creation of a symlink is not implemented for Windows yet, assertions
that consider this state were added when the tests are executed in this OS.
* move is_dyn_import argument from Loader::resolve to Loader::load - it was always kind of strange that resolve() checks permissions.
* change argument type from &str to &ModuleSpecifier where applicable
Ref #3712. This change allowed the deno_typescript crate to reference
cli/js/lib.deno_runtime.d.ts which breaks "cargo package". We intend to
reintroduce a revised version of this patch later once "cargo
package" is working and tested.
This reverts commit 737ab94ea1.
* split ops/worker.rs into ops/worker_host.rs and ops/web_worker.rs
* refactor js/workers.ts and factor out js/worker_main.ts - entry point for WebWorker runtime
* BREAKING CHANGE: remove support for blob: URL in Worker
* BREAKING CHANGE: remove Deno namespace support and noDenoNamespace option in Worker constructor
* introduce WebWorker struct which is a stripped down version of cli::Worker
After landing #3358 the benchmarks exploded indicating problems with workers and deno_core_http_bench.
This PR dramatically fixes thread/syscall count that showed up on benchmarks. Thread count is not back to previous levels but difference went from hundreds/thousands to about ~50.
- removes global `RESOURCE_TABLE` - resource tables are now created per `Worker`
in `State`
- renames `CliResource` to `StreamResource` and moves all logic related
to it to `cli/ops/io.rs`
- removes `cli/resources.rs`
- adds `state` argument to `op_read` and `op_write` and consequently adds
`stateful_minimal_op` to `State`
- IMPORTANT NOTE: workers don't have access to process stdio - this is
caused by fact that dropping worker would close stdout for process
(because it's constructed from raw handle, which closes underlying file
descriptor on drop)
This patch does not work with the recent bundler changes (#3325).
Unfortunately I didn't merge master before landing this patch. It has
something to do with console.log not working inside the compiler worker.
This reverts commit fd62379eaf.
- removes global `RESOURCE_TABLE` - resource tables are now created per `Worker`
in `State`
- renames `CliResource` to `StreamResource` and moves all logic related
to it to `cli/ops/io.rs`
- removes `cli/resources.rs`
- adds `state` argument to `op_read` and `op_write` and consequently adds
`stateful_minimal_op` to `State`
- IMPORTANT NOTE: workers don't have access to process stdio - this is
caused by fact that dropping worker would close stdout for process
(because it's constructed from raw handle, which closes underlying file
descriptor on drop)
Towards simplifying (or better removing entirely) the CoreResource
trait. Resources should be any bit of privileged heap allocated memory
that needs to be referenced from JS, not very specific trait
implementations. Therefore CoreResource should be pushed towards being
as general as possible.
* Split ThreadSafeState into State and GlobalState. State is a "local"
state belonging to "Worker" while "GlobalState" is state shared by
whole program.
* Update "Worker" and ops to use "GlobalState" where applicable
* Move and refactor "WorkerChannels" resource
Basically this does pre-processing of TypeScript files and gathers all the
dependencies asynchronously. Only then after all the dependencies are gathered,
does it do a compile, which at that point all the dependencies are cached in
memory in the compiler, so with the exception of the hard coded assets, there
are no ops during the compilation.
Because op_fetch_source_files is now handled asynchronously in the runtime, we
can eliminate the tokio_util::block_on() which was causing the increase in
threads. Benchmarking on my machine has shown about a 5% improvement in speed
when dealing with compiling TypeScript. Still a long way to go, but an
improvement.
In theory the module name resolution and the fetching of the source files could
be broken out as two different ops. This would prevent situations of sending the
full source file all the time when actually the module is the same module
referenced by multiple modules, but that could be done subsequently to this.
When the global timer fires earlier than expected, which apparently
happens sometimes on server editions of Windows, we didn't call any
setTimeout callbacks, but we *also* didn't reschedule the global timer
to fire again later.
When this situation occurred it would make deno exit abruptly if there
were no other asynchronous ops running on the event loop. It could also
lead to application hangs if the upcoming setTimeout callback was
critical for the application to make progress.
