Cycle detected when optimizing MIR

Code

I tried this code:
https://github.com/FlixCoder/fhir-sdk

cargo build -p fhir-model --release

(Probably the same as creating a crate with dependency fhir-model from crates.io).

The example crate is huge and difficult to navigate, so I tried to reproduce with a minimal example with the same cycle:

#[derive(Debug, PartialEq)]
pub struct Bundle(pub Box<BundleInner>);

#[derive(Debug, PartialEq)]
pub struct BundleInner {
	pub issues: Option<Resource>,
	pub entry: Vec<Option<BundleEntry>>,
}

#[derive(Debug, PartialEq)]
pub struct BundleEntry {
	pub resource: Option<Resource>,
}

#[derive(Debug, PartialEq)]
pub enum Resource {
	Bundle(Bundle),
}

However, this successfully compiles.. So I don't really know how to proceed from here, I am open to suggestions what we could try to find a better small example.

I expected to see this happen: Code compiles with opt-level > 1

Instead, this happened: Compilation error on opt-level > 1

(See FlixCoder/fhir-sdk#56 )

Version it worked on

It most recently worked on: 1.89

Version with regression

Rust 1.90 does not and rustc 1.92.0-nightly (975e6c8 2025-09-23) also does not work.

rustc --version --verbose:

rustc 1.92.0-nightly (975e6c8fe 2025-09-23)
binary: rustc
commit-hash: 975e6c8fec280816d24fbde6b8dfe19620f2efe6
commit-date: 2025-09-23
host: x86_64-unknown-linux-gnu
release: 1.92.0-nightly
LLVM version: 21.1.1

Backtrace

No crash, no backtrace.

Since I have found it in other issues of this kind, I have run cargo bisect-rustc:

cargo bisect-rustc --start=2025-05-25 -- build -p fhir-model --release

It found the following:

********************************************************************************
Regression in nightly-2025-06-29
********************************************************************************

looking for regression commit between 2025-06-28 and 2025-06-29
fetching (via remote github) commits from max(bdaba05a953eb5abeba0011cdda2560d157aed2e, 2025-06-26) to 11ad40bb839ca16f74784b4ab72596ad85587298
ending github query because we found starting sha: bdaba05a953eb5abeba0011cdda2560d157aed2e
get_commits_between returning commits, len: 5
  commit[0] 2025-06-27: Auto merge of #143064 - flip1995:clippy-subtree-update, r=GuillaumeGomez
  commit[1] 2025-06-27: Auto merge of #143116 - matthiaskrgr:rollup-zy9ez06, r=matthiaskrgr
  commit[2] 2025-06-28: Auto merge of #141759 - 1c3t3a:discriminants-query, r=saethlin
  commit[3] 2025-06-28: Auto merge of #142317 - compiler-errors:perf-fold, r=lcnr
  commit[4] 2025-06-28: Auto merge of #142625 - cjgillot:inline-nocycle, r=oli-obk
validated commits found, specifying toolchains

********************************************************************************
Regression in 11ad40bb839ca16f74784b4ab72596ad85587298
********************************************************************************

Attempting to search unrolled perf builds
ERROR: couldn't find perf build comment

searched nightlies: from nightly-2025-05-25 to nightly-2025-09-24
regressed nightly: nightly-2025-06-29
searched commit range: https://github.com/rust-lang/rust/compare/bdaba05a953eb5abeba0011cdda2560d157aed2e...11ad40bb839ca16f74784b4ab72596ad85587298
regressed commit: https://github.com/rust-lang/rust/commit/11ad40bb839ca16f74784b4ab72596ad85587298

<details>
<summary>bisected with <a href='https://github.com/rust-lang/cargo-bisect-rustc'>cargo-bisect-rustc</a> v0.6.10</summary>


Host triple: x86_64-unknown-linux-gnu

@cjgillot

#142625 @cjgillot

We may need help reducing this 29MB source code 😓

Another reduction with expanded derives and simplified output: (requires two library crates foo and x, foo depends on x)

// crate x
pub trait Compare {
    fn eq(self);
}

pub fn wrap<A: Compare>(a: A) {
    Compare::eq(a);
}

// crate foo
use x::{Compare, wrap};

pub struct BundleInner;

impl Compare for BundleInner {
    fn eq(self) {
        lots_of_calls();
        wrap(Resource::ExtensionValue);
    }
}

pub enum Resource {
    Bundle,
    ExtensionValue,
}

impl Compare for Resource {
    fn eq(self) {
        match self {
            Self::Bundle => wrap(BundleInner),
            Self::ExtensionValue => lots_of_calls(),
        }
    }
}

macro_rules! units {
    ($($n: ident)*) => {
        $(
            struct $n;

            impl Compare for $n {
                fn eq(self) {  }
            }

            wrap($n);
        )*
    };
}

fn lots_of_calls() {
    units! {
        O1 O2 O3 O4 O5 O6 O7 O8 O9 O10 O11 O12 O13 O14 O15 O16 O17 O18 O19 O20
        O21 O22 O23 O24 O25 O26 O27 O28 O29 O30 O31 O32 O33 O34 O35 O36 O37 O38 O39 O40
        O41 O42 O43 O44 O45 O46 O47 O48 O49 O50 O51 O52 O53 O54 O55 O56 O57 O58 O59 O60
        O61 O62 O63 O64 O65
    }
}

@rustbot

Assigning priority (discussion on Zulip).

@rustbot label -I-prioritize +P-medium

Auto merge of #147361 - saethlin:fallible-cycle-detection, r=oli-obk

Fixed by #147361.

I am not 100% sure that PR fixed the original issue. Don't hesitate to file a new issue if you run into the problem again.

Just tried the current nightly and it succeeds compiling, thank you :)

Navigation Menu

Repository navigation

Cycle detected when optimizing MIR #146998

Code

Version it worked on

Version with regression

Backtrace

1 remaining item

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Cycle detected when optimizing MIR #146998

Description

Code

Version it worked on

Version with regression

Backtrace

Activity

FlixCoder commented on Sep 24, 2025

Jules-Bertholet commented on Sep 24, 2025

lqd commented on Sep 26, 2025

FlixCoder commented on Sep 27, 2025

cyrgani commented on Sep 27, 2025

1 remaining item

cyrgani commented on Sep 28, 2025

apiraino commented on Oct 29, 2025

cyrgani commented on Jan 22, 2026

saethlin commented on Jan 22, 2026

FlixCoder commented on Jan 24, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions