lcm on i64 falls victim to overflow checking

[pnkfelix]

#[test]
fn demo_i64_lcm_problem() {
    use num::integer::Integer;
    let x: i64 = 46656000000000000;
    let y: i64 = 600;
    let l = x.lcm(&y);
    println!("x: {} y: {} l: {}", x, y, l);
}

This signals overflow for me when run in a debug build:

% RUST_BACKTRACE=1 cargo test demo_i64_lcm
   Compiling qcon-london2016 v0.1.0 (file:///Users/fklock/Dev/Mozilla/pnk-presents/qcon-london2016)
     Running target/debug/lib-d106592ea2cd2eeb

running 0 tests

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

     Running target/debug/mandelbrot-6ba1a3372be98e2a

running 1 test
test demo_i64_lcm_problem ... FAILED

failures:

---- demo_i64_lcm_problem stdout ----
    thread 'demo_i64_lcm_problem' panicked at 'arithmetic operation overflowed', /Users/fklock/.multirust/toolchains/nightly/cargo/registry/src/github.com-88ac128001ac3a9a/num-0.1.30/src/integer.rs:262
stack backtrace:
   1:        0x10dc18e58 - sys::backtrace::tracing::imp::write::h6de1146e783ad759jcu
   2:        0x10dc1a915 - panicking::default_handler::_$u7b$$u7b$closure$u7d$$u7d$::closure.42839
   3:        0x10dc1a477 - panicking::default_handler::hcdc388abac3dab00txy
   4:        0x10dc0ee26 - sys_common::unwind::begin_unwind_inner::h92bb9fa816635994Bbt
   5:        0x10dc0f29e - sys_common::unwind::begin_unwind_fmt::h74474f4da0682187Hat
   6:        0x10dc182f7 - rust_begin_unwind
   7:        0x10dc3c860 - panicking::panic_fmt::hd3ab350bce4886e96YL
   8:        0x10dc3cb5c - panicking::panic::he1e34a74b1285647DXL
   9:        0x10dbb0f85 - integer::i64.Integer::lcm::h95d9e21688ecbb32EKf
  10:        0x10dbb0dbd - demo_i64_lcm_problem::h91d8aeddf6ee4d34tdc
  11:        0x10dbf1d5b - boxed::F.FnBox<A>::call_box::h14880339069920088140
  12:        0x10dbf41cd - sys_common::unwind::try::try_fn::h15106091318301828466
  13:        0x10dc1828b - __rust_try
  14:        0x10dc15f63 - sys_common::unwind::inner_try::he3a51540743a3631D8s
  15:        0x10dbf45b4 - boxed::F.FnBox<A>::call_box::h1300029706881708100
  16:        0x10dc19b9f - sys::thread::Thread::new::thread_start::h26f0f802585e3c80ULx
  17:     0x7fff91138c12 - _pthread_body
  18:     0x7fff91138b8f - _pthread_start


failures:
    demo_i64_lcm_problem

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured

%

[cuviper]

Ah, I see. It currently does `self*other/gcd`, so that multiplication will overflow. Doing the division first should solve this, and we know this is sound since GCD is a divisor. Care to try that and add your test case? (And of course an LCM that's truly too large will still overflow.)

[pnkfelix]

Oh yes, the gcd should indeed divide either one. Okay I'll try that.