Test, should_panic: somehow hide stack traces?

#1

I have a test that is marked with

    #[test]
    #[should_panic]

as a result, in the console output, there is a partial stack trace (which my IDE interprets and makes the file names clickable & blue); as a result, it takes me a while to go from

"something went wrong, I got a stack trace" to "oh, I'm supposed to get a stack trace, this is should_panic"

Question: is there some config flag which tells should_panic to also surpress the stack trace output?

#2

You can override the default panic hook inside the function itself to suppress backtrace printing:

std::panic::set_hook(Box::new(drop));
1 Like
#3

I am getting a very weird error:

254 |         std::panic::set_hook(Box::new(drop));
    |                              ^^^^^^^^^^^^^^ one type is more general than the other
    |
    = note: expected type `FnOnce<(&PanicInfo<'_>,)>`
               found type `FnOnce<(&PanicInfo<'_>,)>`

Especially weird as the two types, as far as I can tell, as identical.

1 Like
#4 
        let f  = |_: &PanicInfo| {};
        std::panic::set_hook(Box::new(f));

appears to work, so you have definitely solved my problem. However, if anyone knows the answer, I'm still curious about the error above.

#5

Ah, it's because Rust doesn't currently have the ability to convert a function generic over a type to a function generic over a lifetime: drop implements for<T> Fn(T), but Rust doesn't see that as a superset of for<'a, 'b> Fn(&'a PanicInfo<'b>), even though it in theory is. So what Rust is doing is instantiating a drop::<&'1 PanicInfo<'2>> where '1 and '2 are some arbitrarily chosen concrete lifetimes, but that doesn't compile because set_hook requires the function to accept a generic lifetime.

1 Like
#6

might as well do

std::panic::set_hook(Box::new(|_| {}))

then :wink:

3 Likes