Bogus error when creating for an impl for a nonexistent type

[jdm]

impl B {
}

fn main() {
}

/tmp/foopy.rs:1:5: 1:6 error: found value name used as a type: def_mod({crate: 0, node: 3})
/tmp/foopy.rs:1 impl B {
                     ^

[owaspjocur]

Could this be fixed by "::unwrap" in the file (src/test/compile-fail/trait-or-new-type-instead.rs)

// error-pattern: found value name used as a type
impl Option::unwrap {
pub fn foo(&self) {

}

fn main() { }

[jdm]

@owaspjocur I don't understand your question, and that's not valid syntax. You can only create an impl for a type in the same module.

[owaspjocur]

definitely I'm confused :-P. Im trying to determine how to fix this issue.Has been a while I didnt work with C like language(last 5 years have been working with C# and Java) ;-)
If I create my own file (as you did writing the foopy.rs) and implement the same code, I get the issue. Rust is a compiler, so Im assuming that a compiler file should be fixed in order to avoid the user implement namespace overlapping imports. Is my assumption correct about this problem?

[jdm]

Yes, your assumption is correct. The goal is to provide an error message that makes sense when a program tries to create an impl for a type that is not in scope (ie. does not exist).

[owaspjocur]

The code I provided is a Rust compiler file (which i changed), not a new one. Im trying to understand where in the code can I implement this fix

[jdm]

Find the place that generates the error message, and follow the code backwards until it's clear where you could modify it to produce a better error message for this case.

[owaspjocur]

Someone gave me a good tip on how to debug Rust http://michaelwoerister.github.io/2013/09/27/what-you-call-the-present.html. I will try this. The terminal is quite archaic ;-P

[huonw]

@owaspjocur I think @jdm's point was essentially "grep for the error message, and look at the code there", using a full debugger to find out where it is is probably still quite hard at the moment.

In any case, running git grep "found name value used as type" turns up this code here. So it looks like one could just expand the _ pattern to cover more cases.

[owaspjocur]

@huonw I used Eclipse as explained by Michael and did a search and found that file. Now correct me if I'm wrong but the it seems that there is no
"match a_def {" (startting at line 444)
for the impl type.

I'm going to attempt to add a case and test

[owaspjocur]

@jdm in ast.rs there is an enumeration, I dont see one for 'impl' (such as DefTrait for example)
line 226.
Isn't suppose to be one?

[jdm]

No, because the name we are creating an impl for is not itself an impl; the compiler seems to think that it is a module instead. Figuring out why this is the case would be a good start.

[owaspjocur]

@jdm seems like astriranna solved the issue by adding:
ast::DefMod(_) =>

      tcx.sess.span_fatal(ast_ty.span,

                          format!("found module name used as a type: {:?}", a_def));
      }