Type inference stopping short of success

[alexcrichton]

For this program

use core::hashmap::HashMap;

fn main() {
  let mut m = HashMap::new();

  match m.find_mut(&1) {
    Some(a) => { a.push(2); }
    None => ()
  }
  m.insert(3, ~[3]);
}

You get the error

foo.rs:7:17: 7:18 error: the type of this value must be known in this context
foo.rs:7     Some(a) => { a.push(2); }
                          ^
error: aborting due to previous error

If the m.insert line were farther up, the program compiles just fine. It seems like if the compiler were to look ahead a bit farther it could figure out the type of the hash map.

[catamorphism]

Seems like a typechecker bug to me (it should gather all the constraints before reporting this error). Though @nikomatsakis can correct me if I'm wrong.

Nominating for milestone 5, production-ready.

[nikomatsakis]

This is working as expected. We use a hybrid form of HM which sometimes requires type information to proceed. Certain operators (the . operator and the * operator, in particular, though perhaps a few others) are overloaded based on the type of the argument and simply cannot be typed otherwise. Sorry, you'll just have to add an annotation on m.