Using `error` as variable name in a pattern

[SimonSapin]

So, I just wasted an hour pulling my hair out about this. Minimal program:

fn main(){
    let result = Some(~"Something went wrong!");
    match copy result {
        Some(error) => io::println(error),
        None => ()
    }
}

rustc version: 09bb07b (master from a few weeks ago)
rustc output:

fuuu.rs:4:35: 4:40 error: mismatched types: expected `&/str` but found `u32` (expected &/str but found u32)
fuuu.rs:4         Some(error) => io::println(error),
                                             ^~~~~
error: aborting due to previous error

This error message is terrible. What’s going on here is that error is a constant from libcore.

The message when using the name of a constant for a new variable is not much better:

fuuu.rs:2:8: 2:15 error: only refutable patterns allowed here
fuuu.rs:2     let error = 4;
                  ^~~~~~~
error: aborting due to previous error

Shouldn’t scoping take care of this? Variables should be able to "mask" constants as long as they’re in scope.

[SimonSapin]

Related to #3941, but even if that is fixed I’d like scoping to allow variables (in let or in a match pattern) to mask constants. If this is decided against, at least have better error messages.

[catamorphism]

It's intentional that variables can't shadow constants. I suppose there's a possibility to change that, but it's not something we've talked about changing.

However, I agree that the error is very bad in this case. Ideally, the error would say something like "there is a constant in scope called error with type u32, you probably meant to introduce a new variable..."

[osaut]

FWIW, the code now works as expected on incoming f54adca.

% rust run test.rs                                                          
warning: no debug symbols in executable (-arch x86_64)
Something went wrong!

[graydon]

fixed in cb37d09

[brendanzab]

Yeah I ran into this only refutable patterns allowed here error today too. I'm guessing it's still hanging around in stage0?