ICE: OutputTypeParameterMismatch #117951

dwhjames · 2023-11-15T21:18:03Z

Error output

Backtrace

error: internal compiler error: compiler/rustc_traits/src/codegen.rs:43:13: Encountered error `
    OutputTypeParameterMismatch(
      SelectionOutputTypeParameterMismatch {
        found_trait_ref: Binder { value: <fn(<round_digest::RoundDigestTag<<Dps as mpc_signing::protocols::cggmp21::DpsProtocol>::SendAllReport2> as mpc_signing::utils::VerifyDeserialize>::VerifyError) -> ds::VerifyError<Dps, Ds> {ds::VerifyError::<Dps, Ds>::Round2} as std::ops::FnOnce<(<round_digest::RoundDigestTag<<Dps as mpc_signing::protocols::cggmp21::DpsProtocol>::SendAllReport2> as mpc_signing::utils::VerifyDeserialize>::VerifyError,)>>, bound_vars: [] },
        expected_trait_ref: Binder { value: <fn(<round_digest::RoundDigestTag<<Dps as mpc_signing::protocols::cggmp21::DpsProtocol>::SendAllReport2> as mpc_signing::utils::VerifyDeserialize>::VerifyError) -> ds::VerifyError<Dps, Ds> {ds::VerifyError::<Dps, Ds>::Round2} as std::ops::FnOnce<(round_digest::VerifyError<<<Dps as mpc_signing::protocols::cggmp21::DpsProtocol>::SendAllReport2 as mpc_signing::utils::VerifyDeserialize>::VerifyError>,)>>, bound_vars: [] },
        terr: Sorts(ExpectedFound {
          expected: round_digest::VerifyError<Alias(Projection, AliasTy { args: [Alias(Projection, AliasTy { args: [Dps/#0], def_id: DefId(22:4767 ~ mpc_signing[4ddb]::protocols::cggmp21::dps_protocol::traits::DpsProtocol::SendAllReport2) })], def_id: DefId(22:3108 ~ mpc_signing[4ddb]::utils::zkp_stream::VerifyDeserialize::VerifyError) })>,
          found: Alias(Projection, AliasTy { args: [round_digest::RoundDigestTag<Alias(Projection, AliasTy { args: [Dps/#0], def_id: DefId(22:4767 ~ mpc_signing[4ddb]::protocols::cggmp21::dps_protocol::traits::DpsProtocol::SendAllReport2) })>], def_id: DefId(22:3108 ~ mpc_signing[4ddb]::utils::zkp_stream::VerifyDeserialize::VerifyError) }) }) }
    )` selecting `
      <fn(<round_digest::RoundDigestTag<<Dps as mpc_signing::protocols::cggmp21::DpsProtocol>::SendAllReport2> as mpc_signing::utils::VerifyDeserialize>::VerifyError)
        -> ds::VerifyError<Dps, Ds> {ds::VerifyError::<Dps, Ds>::Round2} as std::ops::FnOnce<(round_digest::VerifyError<<<Dps as mpc_signing::protocols::cggmp21::DpsProtocol>::SendAllReport2 as mpc_signing::utils::VerifyDeserialize>::VerifyError>,)>>
  ` during codegen

Capture of Nightly failure (2023-10-31) ❌
rustc-ice-2023-10-31T16_49_05-19815.txt

Capture of Beta failure (2023-11-15) ❌
rustc-ice-2023-11-15.txt

Code

From 2023-10-24 to today the source code has not changed. The code is not public, but could potentially be shared. Unfortunately, the code is complex due to the use of a declarative macro, so we do not currently have a minimization.

We would welcome some hints/suggestions about what code/context we could capture to improve this issue.

The text was updated successfully, but these errors were encountered:

lqd · 2023-11-15T22:19:53Z

A bisection using cargo-bisect-rustc (documentation and tutorials are available here; also: use the --access github flag, otherwise it will clone this repo) would at least show the PRs that:

caused the error on nightly, and that PR should now be on beta
fixed the issue on nightly only

It's likely the PR you will find fixing the issue on nightly will also need to be backported to beta.

dwhjames · 2023-11-15T22:55:37Z

@lqd thanks for the directions. I will take a crack at that.

Could you help me understand the release process… will there be any further promotions of changes in nightly to beta? Or is beta ‘locked in’ and explicit action is required to address this issue in beta?
I am fine waiting for stabilization, but I’m obviously concerned about this issue eventually showing up in stable.

lqd · 2023-11-15T23:07:03Z

AFAICT beta has branched already and all further promotions will require explicit decisions and actions. But that’s extremely common and business as usual, there’s an entire section of the weekly compiler meeting dedicated to that.

I could be wrong but I don’t believe crater runs have been done for the next beta yet, and that issue may also show up there (it would if your project was on crates.io for example), but even if not: once more details are known about the severity of the issue, how common it can be, how simple the fix is to backport, and so on, all will be taken into account when these decisions are made.

This will all follow naturally once the PRs are identified.

dwhjames · 2023-11-15T23:39:30Z

searched nightlies: from nightly-2023-10-17 to nightly-2023-10-22
regressed nightly: nightly-2023-10-21
searched commit range: 4578435...249624b
regressed commit: 5cee4f3

bisected with cargo-bisect-rustc v0.6.7

Host triple: x86_64-apple-darwin
Reproduce with:

cargo bisect-rustc --start=2023-10-17 --end=2023-10-22 --access github --regress=ice --preserve --prompt -- build --release

dwhjames · 2023-11-15T23:40:31Z

Had a few false starts here until I figured out the … -- build --release to build in release mode (no ICE in debug mode).

dwhjames · 2023-11-15T23:57:20Z

run again with --regress=non-ice (as this failure PASSED in our weekly build on 2023-10-17 and 2023-10-24, so the failure flip flopped during that period)

searched nightlies: from nightly-2023-10-21 to nightly-2023-10-24
fixed nightly: nightly-2023-10-23
searched commit range: 1c05d50...54b0434
fixed commit: 54b0434

bisected with cargo-bisect-rustc v0.6.7

Host triple: x86_64-apple-darwin
Reproduce with:

cargo bisect-rustc --start=2023-10-21 --end=2023-10-24 --access github --regress=non-ice --preserve --prompt -- build --release

dwhjames · 2023-11-16T00:06:38Z

Further searching 2023-10-24 to 2023-10-31

…
looking for regression commit between 2023-10-25 and 2023-10-26
fetching (via remote github) commits from max(df871fbf053de3a855398964cd05fadbe91cf4fd, 2023-10-23) to ab5c841a1f3c09edc5ea07722519627c960aed17
ending github query because we found starting sha: df871fbf053de3a855398964cd05fadbe91cf4fd
get_commits_between returning commits, len: 13
…

checking the start range to verify it passes
installing df871fbf053de3a855398964cd05fadbe91cf4fd
…

ERROR: the commit at the start of the range (df871fbf053de3a855398964cd05fadbe91cf4fd) includes the regression

bisected with cargo-bisect-rustc v0.6.7

Host triple: x86_64-apple-darwin
Reproduce with:

cargo bisect-rustc --start=2023-10-24 --end=2023-10-31 --access github --regress=ice --preserve -- build --release--release

lqd · 2023-11-16T00:07:42Z

Kind of unexpected, the PR that fixed the issue is about cargo and shouldn’t change how the compiler is called or behaves. Try double checking maybe, preferably with a script looking for the exact ICE message here rather than with --prompt.

dwhjames · 2023-11-16T00:25:27Z

❯ rustup override set nightly-2023-11-14
❯ cargo build --release -p mpc-fsm
💥 

❯ rustup override set nightly-2023-11-15
❯ cargo build --release -p mpc-fsm
✅ 

❯ rustup override unset

Yet

cargo bisect-rustc --start=2023-11-14 --end=2023-11-15 --access github --regress=non-ice --preserve -- build --release

ICE at both start and end 🤔

dwhjames · 2023-11-16T00:45:36Z

repeating that with commit hashes

nightly-2023-11-14-x86_64-apple-darwin unchanged - rustc 1.76.0-nightly (ba7c7a3 2023-11-13)
nightly-2023-11-15-x86_64-apple-darwin unchanged - rustc 1.76.0-nightly (dd430bc 2023-11-14)

cargo bisect-rustc --start=ba7c7a301 --end=dd430bc8c --access github --regress=non-ice --preserve -- build --release -p mpc-fsm

still ICE at both start and end 🤯

dwhjames · 2023-11-16T02:04:41Z

reproducing the search for the first regression within a wider date range

searched nightlies: from nightly-2023-10-01 to nightly-2023-10-22
regressed nightly: nightly-2023-10-21
searched commit range: 4578435...249624b
regressed commit: 5cee4f3

bisected with cargo-bisect-rustc v0.6.7

Host triple: x86_64-apple-darwin
Reproduce with:

cargo bisect-rustc --start=2023-10-01 --end=2023-10-22 --access github --regress=ice --preserve -- build --release -p mpc-fsm

reproducing the return to non-ICE

searched nightlies: from nightly-2023-10-21 to nightly-2023-10-24
fixed nightly: nightly-2023-10-23
searched commit range: 1c05d50...54b0434
fixed commit: 54b0434

bisected with cargo-bisect-rustc v0.6.7

Host triple: x86_64-apple-darwin
Reproduce with:

cargo bisect-rustc --start=2023-10-21 --end=2023-10-24 --access github --regress=non-ice --preserve -- build --release -p mpc-fsm

reproducing the search for the second occurrence of the regression within a wider date range

searched nightlies: from nightly-2023-10-23 to nightly-2023-11-15
regressed nightly: nightly-2023-10-25
searched commit range: f1a5ce1...df871fb
regressed commit: 271dcc1

bisected with cargo-bisect-rustc v0.6.7

Host triple: x86_64-apple-darwin
Reproduce with:

cargo bisect-rustc --start=2023-10-23 --end=2023-11-15 --access github --regress=ice --preserve -- build --release -p mpc-fsm

lqd · 2023-11-16T07:52:28Z

The latter, #116435, is again a PR that shouldn't have any effect in theory: it's a change to the new trait solver, and that is not used unless a -Ztrait-solver=next flag is passed. Backporting this PR shouldn't fix this issue.

We'd need to verify with a way to reproduce this issue: I'll mark this as needing an MCVE, but we don't need that reproducer to be minimal just yet. People can help reducing it afterwards.

dwhjames · 2023-11-16T18:44:49Z

An ideas as to why I might see

❌ nightly-2023-11-14-x86_64-apple-darwin unchanged - rustc 1.76.0-nightly (ba7c7a3 2023-11-13)
✅ nightly-2023-11-15-x86_64-apple-darwin unchanged - rustc 1.76.0-nightly (dd430bc 2023-11-14)
❌ nightly-2023-11-16-x86_64-apple-darwin unchanged - rustc 1.76.0-nightly (6b771f6 2023-11-15)

when using a toolchain override with rustup…

yet, I can not reproduce the build passing with cargo-bisect-rustc? (#117951 (comment))

lqd · 2023-11-16T20:10:25Z

Not easily without being able to reproduce the issue, unfortunately.

dwhjames · 2023-12-14T17:34:27Z

I don’t yet have a minimized version of the issue, but I do have new information.

There are several crates in the project. The crash occurs within mpc-fsm and code generated from a large macro. I have gotten inconsistent results when attempting to compile just mpc-fsm crate (cargo build --release -p mpc-fsm). However, I have found when investigating, that compiling a second crate mpc-jni (cargo build --release -p mpc-jni) that depends on mpc-fsm will successfully compile, even when that same compiler version will crash when compiling mpc-fsm in isolation.

Could it be the case that the compilation context would cause some specialization or ‘guidance’, even across crates?

Does anything about this (the macros or the cross crate behavior) help with how I might attempt to minimize this issue (which is daunting given the size)?

The ‘beta’ channel (now at rustc 1.75.0-beta.5 (1a06ac5b5 2023-12-01)) still has the ICE for this project.

The following is a redo of hunting for the original regression, but compiling with the context of the second crate, which gives a different answer compared to #117951 (comment)

searched nightlies: from nightly-2023-10-17 to nightly-2023-10-22
regressed nightly: nightly-2023-10-19
searched commit range: 09df610...0039d73
regressed commit: 862bba6

bisected with cargo-bisect-rustc v0.6.7

Host triple: x86_64-apple-darwin
Reproduce with:

cargo bisect-rustc --start=2023-10-17 --end=2023-10-22 --access github --regress=ice --preserve -- build --release -p mpc-jni

kpreid · 2023-12-25T15:57:45Z

Relabeling issues which don't have a public runnable reproduction (as opposed to not having a minimized one) to the new label S-needs-repro.
@rustbot label +S-needs-repro -E-needs-mcve

dwhjames added C-bug Category: This is a bug. I-ICE Issue: The compiler panicked, giving an Internal Compilation Error (ICE) ❄️ T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Nov 15, 2023

rustbot added the needs-triage This issue may need triage. Remove it if it has been sufficiently triaged. label Nov 15, 2023

saethlin added E-needs-bisection Call for participation: This issue needs bisection: https://github.com/rust-lang/cargo-bisect-rustc and removed needs-triage This issue may need triage. Remove it if it has been sufficiently triaged. labels Nov 15, 2023

lqd added the E-needs-mcve Call for participation: This issue has a repro, but needs a Minimal Complete and Verifiable Example label Nov 16, 2023

rustbot added S-needs-repro Status: This issue has no reproduction and needs a reproduction to make progress. and removed E-needs-mcve Call for participation: This issue has a repro, but needs a Minimal Complete and Verifiable Example labels Dec 25, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ICE: OutputTypeParameterMismatch #117951

ICE: OutputTypeParameterMismatch #117951

dwhjames commented Nov 15, 2023

lqd commented Nov 15, 2023

dwhjames commented Nov 15, 2023

lqd commented Nov 15, 2023 •

edited

Loading

dwhjames commented Nov 15, 2023

dwhjames commented Nov 15, 2023

dwhjames commented Nov 15, 2023

dwhjames commented Nov 16, 2023

lqd commented Nov 16, 2023

dwhjames commented Nov 16, 2023

dwhjames commented Nov 16, 2023

dwhjames commented Nov 16, 2023

lqd commented Nov 16, 2023

dwhjames commented Nov 16, 2023

lqd commented Nov 16, 2023

dwhjames commented Dec 14, 2023 •

edited

Loading

kpreid commented Dec 25, 2023

ICE: OutputTypeParameterMismatch #117951

ICE: OutputTypeParameterMismatch #117951

Comments

dwhjames commented Nov 15, 2023

Meta

Error output

Code

lqd commented Nov 15, 2023

dwhjames commented Nov 15, 2023

lqd commented Nov 15, 2023 • edited Loading

dwhjames commented Nov 15, 2023

dwhjames commented Nov 15, 2023

dwhjames commented Nov 15, 2023

dwhjames commented Nov 16, 2023

lqd commented Nov 16, 2023

dwhjames commented Nov 16, 2023

dwhjames commented Nov 16, 2023

dwhjames commented Nov 16, 2023

reproducing the search for the first regression within a wider date range

reproducing the return to non-ICE

reproducing the search for the second occurrence of the regression within a wider date range

lqd commented Nov 16, 2023

dwhjames commented Nov 16, 2023

lqd commented Nov 16, 2023

dwhjames commented Dec 14, 2023 • edited Loading

kpreid commented Dec 25, 2023

lqd commented Nov 15, 2023 •

edited

Loading

dwhjames commented Dec 14, 2023 •

edited

Loading