One method calling another crashes rustboot

414b23c
Opened by Jeffrey Yasskin at 2010-07-19 19:05:11

Given:

obj oT() {
  fn get() -> int {
    ret 3;
  }
  fn foo() {
    auto c = get();
  }
}
fn main() {
}

Rustboot crashes with:

Fatal error: exception Failure("internal_check_base_lval: unexpected defn type") Raised at file "pervasives.ml", line 22, characters 22-33 Called from file "boot/me/type.ml", line 280, characters 15-77 Called from file "boot/me/type.ml", line 461, characters 17-47 Called from file "boot/me/type.ml", line 492, characters 6-64 Called from file "boot/me/type.ml", line 598, characters 20-37 Called from file "boot/me/type.ml", line 715, characters 23-45 Called from file "boot/me/type.ml", line 820, characters 8-23 Re-raised at file "boot/me/type.ml", line 633, characters 6-189 Called from file "boot/me/type.ml", line 926, characters 8-51 Called from file "boot/me/walk.ml", line 134, characters 4-9 Called from file "array.ml", line 117, characters 31-48 Called from file "boot/me/walk.ml", line 135, characters 4-15 Called from file "boot/me/walk.ml", line 320, characters 2-23 Called from file "hashtbl.ml", line 145, characters 8-13 Called from file "hashtbl.ml", line 148, characters 4-19 Called from file "boot/me/walk.ml", line 186, characters 12-59 Called from file "boot/me/walk.ml", line 135, characters 4-15 Called from file "hashtbl.ml", line 145, characters 8-13 Called from file "hashtbl.ml", line 148, characters 4-19 Called from file "boot/me/walk.ml", line 135, characters 4-15 Called from file "boot/me/type.ml", line 955, characters 4-231 Re-raised at file "boot/me/type.ml", line 35, characters 4-112 Called from file "boot/driver/main.ml", line 322, characters 9-26 Called from file "array.ml", line 117, characters 31-48 Called from file "boot/driver/main.ml", line 320, characters 4-366 Called from file "boot/driver/main.ml", line 404, characters 5-21

The crash goes away if foo() calls a global function, and I get an appropriate error message if the function is just misspelled.

  1. Bug in the new typechecker and/or resolver; I'll fix that part. It should trap earlier.

    However, this is also a language non-feature (at present). Currently there's no mechanism for self-dispatch inside an object, either implicitly as you're doing, or via this.get(). There's no 'this' binding at all. It's not a ton of work to add, I've been hesitant to see if (a) it's necessary and (b) it will interact with whatever mechanism gets adopted for ad-hoc object extension / method replacement, which we haven't settled on yet.

    And I was also curious to see how far we could get without the feature. In general I only wish to adopt OO features "one at a time" as seems necessary, and stick to the minimum. The "class" concept in many OO languages is the semantic dumping ground of every language feature ever, and I don't want to wind up in that boat.

    Graydon Hoare at 2010-07-17 17:50:52

  2. Ah, didn't realize that. I've documented this restriction in jyasskin/rust@6e967f48958a550fed0c14beda1c40e91356219c. Not having it doesn't seem like the end of the world to me, and that'll certainly simplify the specification of method overriding.

    Jeffrey Yasskin at 2010-07-17 19:21:54

  3. Fix over-optimistic resolution of self-methods within obj scopes. There is no such feature in the language at present. Add test to prevent regression. Closed by f1db420317a53eacf3dfb08b8121ea06ad1ca5b5.

    Graydon Hoare at 2010-07-19 19:05:11