libtest: do not early exit from test runners

[RalfJung]

Suggested by @Mark-Simulacrum in #161868. I finally figured out why my earlier attempts did not work.

[RalfJung]

@bors try job=dist-x86_64-msvc

[rust-bors]

💔 Test for 40346c4 failed: CI. Failed job:

• try - dist-x86_64-msvc (web logs, enhanced plaintext logs)

[rustbot]

clippy is developed in its own repository. If possible, consider making this change to rust-lang/rust-clippy instead.

cc @rust-lang/clippy

[rustbot]

r? @oli-obk

rustbot has assigned @oli-obk.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: compiler
• compiler expanded to 76 candidates
• Random selection from 19 candidates

[RalfJung]

@bors try job=dist-x86_64-msvc,testmsvc

[rust-bors]

☀️ Try build successful (CI)
Build commit: 06e803b (06e803b7911a3f2d76a1a9787f21e39692f30738)
Base parent: 5392d2f (5392d2f545c6836dc79f209bcc14ac7179dbc4f2)

[oli-obk]

@bors r+

[rust-bors]

📌 Commit 5fd598e has been approved by oli-obk

It is now in the queue for this repository.

[ranger-ross]

I think this change has caused Cargo's build-std CI check to start failing on all PRs. (example)

error[E0463]: can't find crate for `test`

It's not clear to me if its an issue with rustc or if Cargo's tests should be updated

[RalfJung]

Hm, that is strange.
Are these tests using the #[test_runner] attribute? Reading the diff now it seems like previously we never imported the test crate if that attribute was set, now we do.

[ranger-ross]

Looking at the failing test, it specifically for custom test harnesses.

#![no_std]
#![cfg_attr(test, no_main)]
#![feature(custom_test_frameworks)]
#![test_runner(crate::test_runner)]

pub fn test_runner(_tests: &[&dyn Fn()]) {}

#[panic_handler]
fn panic(_info: &core::panic::PanicInfo) -> ! {
    loop {}
}

[RalfJung]

Ah, makes sense then.

I suspect adding a #[test] would make it fail even without this PR as that requires extern crate test;...
Also the signature for test_runner seems wrong, we are passing a slice of &TestDescAndFn, not &dyn Fn.

EDIT: Ah, you're supposed to use #[test_case], not #[test], with a custom runner. Then you can pick ~any type signature. #162971 adds a pretty-printing test based on the unstable book so we can at least see that there was an unnecessary extern crate test;.