tests with #[should_panic] should print a message on failure

[fbenkstein]

When a #[should_panic] #[test] fails because there was no panic it should produce an error message to that effect.

For example, the following test:

#[test]
#[should_panic]
fn doesnt_actually_panic() {}

when run with cargo test produces the following output:

test doesnt_actually_panic ... FAILED

failures:

failures:
    doesnt_actually_panic

There is no indication why it failed. Which can be problematic if the test itself is created by a macro. Example: https://github.com/BurntSushi/fst/blob/master/src/raw/tests.rs#L131-L140

[varkor]

I'm not familiar with the code, but I'd guess that you just need to change TrFailed in here:
https://github.com/rust-lang/rust/blob/036e368d6883e0bc141c1b54578876cd5bfb2468/src/libtest/lib.rs#L1525-L1548
to TrFailedMsg("test did not panic as expected").
(And add a test to make sure it works.)

[chansuke]

Thanks.I would like to work on this.

[varkor]

@chansuke: go ahead! If you have any questions, feel free to ask them here.

[chansuke]

@varkor Hi.I've added the TrFailedMsg("test did not panic as expected") in fn calc_result() and trying to write a test but is this right approach?

    #[test]
    fn test_calc_result() {
        let args = // what should I write?
        let panic_message = "test did not panic as expected"
        let result = calc_result(args)
        assert_eq!(result, panic_message)
    }

[varkor]

@chansuke: actually, I think you can get away with just changing this line:

rust/src/libtest/lib.rs

Line 1925 in 7158ed9

assert!(res == TrFailed);

to update it to your TrFailedMsg, because such a test already exists for this case.

[chansuke]

@varkor Thanks

[varkor]

The previous fix was closed, so this issue is open again. The previous PR contains detailed instructions.

[kungfukennyg]

I'd like to finish this out if @chansuke is no longer working on it.