Re-exported concat macro causes "failed to resolve macro" error

[NicholasGorski]

Minimal example, reduced from a larger macro:

pub mod private {
    pub use core::concat;
}

macro_rules! demo {
    ($fmt:expr $(, $($args:tt)*)?) => {{
        // The use of `$crate::private::concat` here causes an error,
        // but unqualified `concat!` does not. They ought to be identical?
        __impl_demo!($crate::private::concat!(": ", $fmt), $($($args)*)?)
    }};
}

macro_rules! __impl_demo {
    ($fmt:expr $(, $($args:tt)*)?) => {{
        // This one doesn't care about which concat is used.
        //
        // However, the use of panic! seems to be necessary; for example,
        // using `println!` works as expected. Weirder, if instead one does
        // `assert!(true, ...)` then *both* expansions below give the error.
        panic!(concat!("prefix", $fmt), $($($args)*)?);
    }};
}

pub fn foo(i: u32) {
    if i == 0 {
        demo!("{}", 1); // A
    }

    if i == 1 {
        demo!(""); // B
    }
}

As is, this is valid code that compiles. However, rust analyzer emits "failed to resolve macro" at expansion B.

Furthermore, as noted one can use println! and neither expansion has an error, or switch to assert! and get an error at both expansions. In all cases there is no actual error during rustc compilation.

(The use of re-exported concat in the private module is to avoid call-site environment pollution in the macro expansion.)

Let me know if this is unclear or if I've made an error along the way. Thanks!

[jonas-schievink]

Reduces to:

mod private {
    pub use core::concat;
}

macro_rules! m {
    () => {
        panic!(concat!($crate::private::concat!("")));
    };
}

fn f() {
    m!();
}

Might be caused by a limitation in our $crate handling (perhaps we're losing information about the origin crate after passing through panic?)

[xffxff]

Can't reproduce it.
RA version: 73668334f05c3446b04116ccc3156240d2d8ab19 with cargo xtask install

[NicholasGorski]

Hm, strange. I am using v0.2.817 (~1 hour ago at time of writing) and still reproducing the issue. Am I incorrectly updated?

[jonas-schievink]

No, I can still repro with both snippets in this thread