The motivation for creating this README and repo was originally from the gnome-class project. This project is defining a procedural macro with a small DSL to conveniently declare glib-compatible types and interface to them from both Rust and C. The crate is currently (as of this writing) undergoing a rewrite to transition to a "new" procedural macro which preserves span information to ensure that type errors and such in user-written code are legible.

So what I did was enable the unstable feature of the proc-macro2 crate (which forces it to implement itself with the upstream proc_macro crate) and then oh boy down the rabbit hole I went...

This project has a few pieces:

• src/lib.rs - this is all the procedural macros for all the tests. There's some more comments about the file internally.
• tests/{failure,success}N.rs - these are a bunch of files showing off whether compilation fails or succeeds. For example failure1.rs is an example of a procedural macro invocation that fails to compile, while success1.rs is a small tweak on behalf of the procedural macro author's side of things which gets the procedural macro to succeed to compile.
• foo - a helper crate for some of the test cases.

For example here's what happened with the failure2 test:

$ cargo +nightly test --test failure2
   Compiling wut v0.1.0 (file:///Users/acrichton/code/weird-proc-macro-spans)
error[E0432]: unresolved import `super`
 --> tests/failure2.rs:8:1
  |
8 | / failure2! {
9 | | }
  | |_^ no `A` in the root

error: aborting due to previous error

error: Could not compile `wut`.

This was an example, if we look at the procedural macro definition, of how a generated module can't import contents defined outside. We can see the fix, however:

$ cargo +nightly test --test success2
    Finished dev [unoptimized + debuginfo] target(s) in 0.0 secs
     Running target/debug/deps/success2-c6ae2bc09f230207

running 0 tests

test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out

One possible fix in this case, looking at the definition of success2, was to switch all spans to call_site.