NodeWorker.sendMessageToWorker fails with multi-byte code points in JSON

[connor4312]

Version

21.1.0

Platform

Darwin mbp.peet.io 23.1.0 Darwin Kernel Version 23.1.0: Mon Oct 9 21:27:27 PDT 2023; root:xnu-10002.41.9~6/RELEASE_X86_64 x86_64

Subsystem

inspector

What steps will reproduce the bug?

From a user's point of view...

1. Debug a script like this in VS Code

const { Worker, isMainThread } = require('worker_threads');

if (isMainThread) {
  new Worker(__filename);
} else {
  debugger;
}

1. Ensure the worker_thread session is select and run console.log('本機磁碟')
2. Nothing happens (an error behind the scenes)

How often does it reproduce? Is there a required condition?

100%

What is the expected behavior? Why is that the expected behavior?

It should log 本機磁碟

What do you see instead?

An error

Additional information

Behind the scenes, this calls

{
  "id": 1027,
  "method": "NodeWorker.sendMessageToWorker",
  "params": {
    "message": "{\"id\":1014,\"method\":\"Debugger.evaluateOnCallFrame\",\"params\":{\"expression\":\"console.log('本機磁碟')\\n//# sourceURL=eval-1e14c9c5.repl\\n\",\"includeCommandLineAPI\":true,\"objectGroup\":\"console\",\"generatePreview\":true,\"callFrameId\":\"-7882104772351236078.1.0\"}}",
    "sessionId": "1"
  }
}

This is created by simply JSON.stringify'ing the relevant CDP call.

However, this results in

{
  "method": "NodeWorker.receivedMessageFromWorker",
  "params": {
    "sessionId": "1",
    "message": "{\"error\":{\"code\":-32700,\"message\":\"Message must be a valid JSON\"}}"
  }
}

Which seems to come from here

node/tools/inspector_protocol/lib/DispatcherBase_cpp.template

Line 234 in 60e8364

reportProtocolErrorTo(m_frontendChannel, DispatchResponse::kParseError, "Message must be a valid JSON");

. I would expect any JSON created by JSON.stringify would be able to be parsed by the NodeWorker domain.

This causes more 'real' issues when users have paths with multi-byte codepoints, as reported in microsoft/vscode-js-debug#1798 and microsoft/vscode-js-debug#2043

[debadree25]

Is there script to reproduce this without having to using the vscode debugger

[connor4312]

This should be pretty easily reproducible by manually interacting with the NodeWorker domain using require('inspector'). The parsing fails when the code point is in any message sent to the worker, no specific patterns or setup is necessary.

[cola119]

Logging English characters works fine.

[cola119]

This issue could be solved by updating our copies of files related to inspector protocol to the latest. I'm working on it.

[connor4312]

This causes programs to break when they use module.register from a filename with multibyte code points (see most recently linked issue)