How to shutdown tokio runtime when handler panics?

[lightsing]

I'm using the 'fail-fast' approach in my code, but I found that panic happens in handler does not cause the whole program shutdown.
Even my tokio runtime is configured using unhandled_panic api in tokio-unstable:

    tokio::runtime::Builder::new_multi_thread()
        .unhandled_panic(UnhandledPanic::ShutdownRuntime)
        .enable_all()
        .build()
        .unwrap()
        .block_on(async_main())

The version of axum I used is rev b80d1261377a59d7a498f1f3b8f8d160e1c61d9e from git.
rustc version: rustc 1.70.0-nightly (0c61c7a97 2023-03-25)
other deps are:

tokio = { version = "1.28", features = ["sync", "macros", "rt-multi-thread"] }
tower-http = { version = "0.4", features = ["full"] }

[Darksonn]

If setting unhandled_panic(UnhandledPanic::ShutdownRuntime) doesn't work, then that implies that someone is catching the panics before they reach Tokio.

[lightsing]

Does axum use catch_unwind when wrapping the handler?

[davidpdrsn]

No but I believed hyper might. Not sure.

[lightsing]

Well, I set panic=abort to resolve this which kind like a hack here