Capture stack trace from main thread

[timfish]

What is the problem this feature will solve?

If synchronous user code blocks the event loop for an extended or indefinite period of time, it would be useful to be able to capture a stack trace to help fix the issue.

Preferably we should be able to capture a stack trace for the main thread from a worker thread so this can be handled in JavaScript.

What is the feature you are proposing to solve the problem?

Improve the diagnostic report feature?

What alternatives have you considered?

Inspector API

You can pause and capture stack traces via the inspector API but it's often not suitable for production usage.

• Debugger can impact performance
• It's not trivial to convert inspector stack frames to the error.stack format
• Serverless embedders can build node without inspector support (Vercel but probably others too!)
• Enabling Node debugger in production has potential security issues

Diagnostic report API

The diagnostic report is almost there but it has a few issues:

• When this feature was brought into core, it didn't retain the watchdog thread that allowed it to interrupt via signal and capture a stack trace from the main thread. The signal handling is now in JavaScript so it is only called if or when the event loop becomes unblocked, by which time the call stack is not relevant. Does anyone have context as to why it was done this way?
• Report is only to file or stdout 😢

The diagnostic report does collect stack traces from worker threads because it interrupts them.