Panicking in a Channel::send closure crashes Node

[kjvalencik]

Repro

use neon::prelude::*;

#[neon::main]
fn main(mut cx: ModuleContext) -> NeonResult<()> {
    cx.export_function("crash", |mut cx| {
        cx.channel().send::<(), _>(|_| panic!("Oh, no!"));
        Ok(cx.undefined())
    })
}

require(".").crash();

Expected

As of Neon 0.10, panicking in a Channel::send closure will trigger in an uncaughtException.

Actual

The process crashes with a C++ error.

> thread '<unnamed>' panicked at 'Oh, no!', test/futures/src/lib.rs:64:40
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
Uncaught:
[Error: A panic occurred while executing a `neon::event::Channel::send` callback] {
  panic: [Error: Oh, no!]
}
> node[25182]: ../src/node_util.cc:242:static void node::util::WeakReference::DecRef(const FunctionCallbackInfo<v8::Value> &): Assertion `(weak_ref->reference_count_) >= (1)' failed.
 1: 0x106604815 node::Abort() (.cold.1) [../node/16.13.2/bin/node]
 2: 0x105303b09 node::Abort() [../node/16.13.2/bin/node]
 3: 0x105303941 node::Assert(node::AssertionInfo const&) [../node/16.13.2/bin/node]
 4: 0x105399dc6 node::util::WeakReference::DecRef(v8::FunctionCallbackInfo<v8::Value> const&) [../node/16.13.2/bin/node]
 5: 0x1054ed4a9 v8::internal::FunctionCallbackArguments::Call(v8::internal::CallHandlerInfo) [../node/16.13.2/bin/node]
 6: 0x1054ecf76 v8::internal::MaybeHandle<v8::internal::Object> v8::internal::(anonymous namespace)::HandleApiCallHelper<false>(v8::internal::Isolate*, v8::internal::Handle<v8::internal::HeapObject>, v8::internal::Handle<v8::internal::HeapObject>, v8::internal::Handle<v8::internal::FunctionTemplateInfo>, v8::internal::Handle<v8::internal::Object>, v8::internal::BuiltinArguments) [../node/16.13.2/bin/node]
 7: 0x1054ec6ef v8::internal::Builtin_HandleApiCall(int, unsigned long*, v8::internal::Isolate*) [../node/16.13.2/bin/node]
 8: 0x105d5d5f9 Builtins_CEntry_Return1_DontSaveFPRegs_ArgvOnStack_BuiltinExit [../node/16.13.2/bin/node]

Explanation

There is a bug in Node that crashes the process when attempting to trigger an uncaughtException with napi_fatal_exception from the context of an error domain.

Since Node-API implicitly wraps napi_threadsafe_function callbacks in an error domain and threadsafe functions are the backing provider of Channel::send, panicking in a neon::event::Channel::send callback crashes the process on all current versions of Node.

Possible Mitigations

• Disable uncaughtException triggering which trades a crash for a silent error
• Try to trigger the exception from outside an error domain. Possibly with setImmediate?
• Wait for a fix from Node

[kjvalencik]

Fixed.