Cranelift: panic calculating timing total

[ConradIrwin]

Steps to Reproduce

Unclear, this is a crash we saw in our telemetry at https://zed.dev. I see that there's a debug_assert that the passes should be in the right order, but in a release build the panic happens later.

overflow when subtracting durations

core::time::impl$3::sub (time.rs:1164)
cranelift_codegen::timing::enabled::impl$0::total::closure$0 (timing.rs:174)
core::ops::function::impls::impl$4::call_once (function.rs:310)
core::option::Option::map (option.rs:1165)
core::iter::adapters::map::impl$2::next (map.rs:107)
core::time::impl$10::sum (time.rs:1259)
core::iter::traits::iterator::Iterator::sum (iterator.rs:3581)
cranelift_codegen::timing::enabled::PassTimes::total (timing.rs:174)
wasmtime_cranelift::compiler::impl$3::compile_function (compiler.rs:284)
wasmtime::compile::impl$4::collect_inputs_in_translations::closure$0 (compile.rs:490)
core::ops::function::FnOnce::call_once<T> (function.rs:250)
alloc::boxed::impl$31::call_once (boxed.rs:2206)
wasmtime::compile::compile_required_builtins::closure$1 (compile.rs:610)
wasmtime::engine::impl$2::run_maybe_parallel::closure$0 (engine.rs:183)
core::ops::function::impls::impl$1::call_mut (function.rs:274)
core::ops::function::impls::impl$4::call_once (function.rs:310)
core::option::Option::map (option.rs:1165)
core::iter::adapters::map::impl$2::next (map.rs:107)
rayon::iter::plumbing::Folder::consume_iter (mod.rs:173)

Expected Results

No panic

Actual Results

Panic

Versions and Environment

Cranelift version or commit: 33.0.2

Operating system: Windows

Architecture: x86

Extra Info

Anything else you'd like to add?

[xyber-nova]

same