This PR implements the child_process IPC pipe between parent and child.
The implementation uses Windows named pipes created by parent and passes
the inheritable file handle to the child.
I've also replace parts of the initial implementation which passed the
raw parent fd to JS with resource ids instead. This way no file handle
is exposed to the JS land (both parent and child).
`IpcJsonStreamResource` can stream upto 800MB/s of JSON data on Win 11
AMD Ryzen 7 16GB (without `memchr` vectorization)
This PR implements the Node child_process IPC functionality in Deno on
Unix systems.
For `fd > 2` a duplex unix pipe is set up between the parent and child
processes. Currently implements data passing via the channel in the JSON
serialization format.
This commit adds granular `--unstable-*` flags:
- "--unstable-broadcast-channel"
- "--unstable-ffi"
- "--unstable-fs"
- "--unstable-http"
- "--unstable-kv"
- "--unstable-net"
- "--unstable-worker-options"
- "--unstable-cron"
These flags are meant to replace a "catch-all" flag - "--unstable", that
gives a binary control whether unstable features are enabled or not. The
downside of this flag that allowing eg. Deno KV API also enables the FFI
API (though the latter is still gated with a permission).
These flags can also be specified in `deno.json` file under `unstable`
key.
Currently, "--unstable" flag works the same way - I will open a follow
up PR that will print a warning when using "--unstable" and suggest to use
concrete "--unstable-*" flag instead. We plan to phase out "--unstable"
completely in Deno 2.
To fix bugs around detection of when node emulation is required, we will
just eagerly initialize it. The improvements we make to reduce the
impact of the startup time:
- [x] Process stdin/stdout/stderr are lazily created
- [x] node.js global proxy no longer allocates on each access check
- [x] Process checks for `beforeExit` listeners before doing expensive
shutdown work
- [x] Process should avoid adding global event handlers until listeners
are added
Benchmarking this PR (`89de7e1ff`) vs main (`41cad2179`)
```
12:36 $ third_party/prebuilt/mac/hyperfine --warmup 100 -S none './deno-41cad2179 run ./empty.js' './deno-89de7e1ff run ./empty.js'
Benchmark 1: ./deno-41cad2179 run ./empty.js
Time (mean ± σ): 24.3 ms ± 1.6 ms [User: 16.2 ms, System: 6.0 ms]
Range (min … max): 21.1 ms … 29.1 ms 115 runs
Benchmark 2: ./deno-89de7e1ff run ./empty.js
Time (mean ± σ): 24.0 ms ± 1.4 ms [User: 16.3 ms, System: 5.6 ms]
Range (min … max): 21.3 ms … 28.6 ms 126 runs
```
Fixes https://github.com/denoland/deno/issues/20142
Fixes https://github.com/denoland/deno/issues/15826
Fixes https://github.com/denoland/deno/issues/20028
This commit changes how data required to bootstrap main and worker
runtime is serialized.
Instead of relying on serde_v8 and using JSON object,
we're doing manual serialization to a "v8::Array". This limits number
of V8 strings that need to be serialized by 16.
It also made it clear that some data could be obtained during
snapshotting instead of during bootstrap.
This commit further improves startup time by:
- no relying on "JsRuntime::execute_script" for runtime bootstrapping,
this is instead done using V8 APIs directly
- registering error classes during the snapshot time, instead of on
startup
Further improvements can be made, mainly around removing
"core.initializeAsyncOps()" which takes around 2ms.
This commit should result in ~1ms startup time improvement.
This adds an implementation of `Default` for both `WorkerOptions` and
`BootstrapOptions`. Since both of these structs are rather big, this
should make it easier for people unfamiliar with the internals to focus
on the options relevant to them.
As a user of `deno_runtime` I feel like these should serve as good
defaults, getting people them started without having to tweak the
runtime. Additionally even if some changes are made, the usage of
`..Default::default()` will significantly help with code clarity and
verbosity.
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.
