[Bug] nested mul_div panicked

[0xbe1]

🐛 Bug

To reproduce

Code snippet to reproduce

This test passes

#[test]
fun test_nested_mul_div() {
    let a = math64::mul_div(1, 1, 1);
    assert!(math64::mul_div(1, a,1) == 1, 0);
}

This test throws exception.

#[test]
fun test_nested_mul_div() {
    assert!(math64::mul_div(1, math64::mul_div(1, 1, 1),1) == 1, 0);
}

error:

thread 'main' panicked at 'called `Option::unwrap()` on a `None` value', third_party/move/move-compiler/src/to_bytecode/context.rs:204:73
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

Expected Behavior

nested should be fine

System information

{
  "Result": {
    "build_branch": "main",
    "build_cargo_version": "cargo 1.66.1 (ad779e08b 2023-01-10)",
    "build_clean_checkout": "false",
    "build_commit_hash": "c9cb442e6f03aaa37687d30899480d40408d0719",
    "build_is_release_build": "true",
    "build_os": "macos-x86_64",
    "build_pkg_version": "2.0.0",
    "build_profile_name": "cli",
    "build_rust_channel": "1.66.1-x86_64-apple-darwin",
    "build_rust_version": "rustc 1.66.1 (90743e729 2023-01-10)",
    "build_tag": "",
    "build_time": "2023-06-01 19:50:01 +00:00",
    "build_using_tokio_unstable": "true"
  }
}

[movekevin]

After some quick debugging, I think this is due to the way inline functions work - you can't have inline functions inside inline functions. cc @wrwg

[wrwg]

After some quick debugging, I think this is due to the way inline functions work - you can't have inline functions inside inline functions. cc @wrwg

Definitely you can have inline functions inside inline functions, it happens frequently in the functionals in the framework. This must be something different.

[movekevin]

Doing this currently fails with the following error:

bug[ICE12001]: BYTECODE GENERATION FAILED
  ┌─ /Users/kevin/aptos-core/aptos-move/framework/aptos-stdlib/sources/math64.move:2:19
  │
2 │ module aptos_std::math64 {
  │                   ^^^^^^ IR ERROR: Unbound function Self.mul_div

[mkurnikov]

@wrwg @movekevin

Definitely you can have inline functions inside inline functions, it happens frequently in the functionals in the framework. This must be something different.

I think it currently requires nested inline function to be defined at the same module as the invocation place.

BYTECODE GENERATION FAILED

The same appears with inline functions that have constants, try to call math128::ceil_div() from anywhere other than math128 itself.

P.S. I suggest Aptos developers to implement tests for the public functions in modules different from the declaration module, i.e. math128_tests.move for math128. It helps catch those bugs early, and makes sure that users have all the necessary APIs to initialize things for example.

[wrwg]

Fixed