Clippy --fix panics with index error

[tvallotton]

Summary

cargo clippy --fix causes an index error.

Reproducer

I pushed the code that causes the error to this repo, so you can reproduce it. The error was triggerd by cargo clippy --fix. Note that this does not seem to panic in stable.

thread 'main' panicked at 'slice index starts at 1 but ends at 0', library/core/src/slice/index.rs:91:5
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
error: could not compile `chess`
warning: build failed, waiting for other jobs to finish...
thread 'main' panicked at 'slice index starts at 1 but ends at 0', library/core/src/slice/index.rs:91:5
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
error: could not compile `chess`

Version

rustc 1.61.0-nightly (76d770ac2 2022-04-02)
binary: rustc
commit-hash: 76d770ac21d9521db6a92a48c7b3d5b2cc535941
commit-date: 2022-04-02
host: aarch64-apple-darwin
release: 1.61.0-nightly
LLVM version: 14.0.0

Additional Labels

No response

[tvallotton]

it looks like its related to #8718.

[xFrednet]

Hey, thank you for the report. This looks like the linked issue. Could you provide the Stacktrace, by setting RUST_BACKTRACE=1? 🙃

[tvallotton]

You can also try to replicate it yourself on this repo. I get this:

thread 'main' panicked at 'slice index starts at 1 but ends at 0', library/core/src/slice/index.rs:91:5
stack backtrace:
   0: _rust_begin_unwind
   1: core::panicking::panic_fmt
   2: core::slice::index::slice_index_order_fail_rt
   3: core::ops::function::FnOnce::call_once
   4: core::intrinsics::const_eval_select
   5: core::slice::index::slice_index_order_fail
   6: rustfix::parse_snippet
   7: rustfix::collect_span
   8: <alloc::vec::Vec<rustfix::Replacement> as alloc::vec::spec_from_iter::SpecFromIter<rustfix::Replacement, core::iter::adapters::filter_map::FilterMap<core::iter::adapters::filter::Filter<core::slice::iter::Iter<rustfix::diagnostics::DiagnosticSpan>, rustfix::collect_suggestions<std::collections::hash::map::RandomState>::{closure#1}::{closure#0}>, rustfix::collect_span>>>::from_iter
   9: <&mut rustfix::collect_suggestions<std::collections::hash::map::RandomState>::{closure#1} as core::ops::function::FnMut<(&rustfix::diagnostics::Diagnostic,)>>::call_mut
  10: <alloc::vec::Vec<rustfix::Solution> as alloc::vec::spec_from_iter::SpecFromIter<rustfix::Solution, core::iter::adapters::filter_map::FilterMap<core::slice::iter::Iter<rustfix::diagnostics::Diagnostic>, rustfix::collect_suggestions<std::collections::hash::map::RandomState>::{closure#1}>>>::from_iter
  11: rustfix::collect_suggestions::<std::collections::hash::map::RandomState>
  12: cargo::ops::fix::fix_maybe_exec_rustc
  13: cargo::main
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
error: could not compile `chess`
warning: build failed, waiting for other jobs to finish...
thread 'main' panicked at 'slice index starts at 1 but ends at 0', library/core/src/slice/index.rs:91:5
stack backtrace:
   0: _rust_begin_unwind
   1: core::panicking::panic_fmt
   2: core::slice::index::slice_index_order_fail_rt
   3: core::ops::function::FnOnce::call_once
   4: core::intrinsics::const_eval_select
   5: core::slice::index::slice_index_order_fail
   6: rustfix::parse_snippet
   7: rustfix::collect_span
   8: <alloc::vec::Vec<rustfix::Replacement> as alloc::vec::spec_from_iter::SpecFromIter<rustfix::Replacement, core::iter::adapters::filter_map::FilterMap<core::iter::adapters::filter::Filter<core::slice::iter::Iter<rustfix::diagnostics::DiagnosticSpan>, rustfix::collect_suggestions<std::collections::hash::map::RandomState>::{closure#1}::{closure#0}>, rustfix::collect_span>>>::from_iter
   9: <&mut rustfix::collect_suggestions<std::collections::hash::map::RandomState>::{closure#1} as core::ops::function::FnMut<(&rustfix::diagnostics::Diagnostic,)>>::call_mut
  10: <alloc::vec::Vec<rustfix::Solution> as alloc::vec::spec_from_iter::SpecFromIter<rustfix::Solution, core::iter::adapters::filter_map::FilterMap<core::slice::iter::Iter<rustfix::diagnostics::Diagnostic>, rustfix::collect_suggestions<std::collections::hash::map::RandomState>::{closure#1}>>>::from_iter
  11: rustfix::collect_suggestions::<std::collections::hash::map::RandomState>
  12: cargo::ops::fix::fix_maybe_exec_rustc
  13: cargo::main
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
error: could not compile `chess`

[Serial-ATA]

Looks like it's coming from the map_flatten lint here: https://github.com/tvallotton/chess/blob/98b01208e95173388cf4f781cc83bb414bb3bf55/src/table.rs#L73

It doesn't like the long multiline suggestions.

For example, this works:

[0u8, 1, 2, 3]
    .into_iter()
    .map(|n| match n {
        1 => [n.saturating_add(1)],
        n => [n],
    })
    .flatten();

While this causes the panic:

[0u8, 1, 2, 3]
    .into_iter()
    .map(|n| match n {
        1 => [n
            .saturating_add(1)
            .saturating_add(1)
            .saturating_add(1)
            .saturating_add(1)
            .saturating_add(1)
            .saturating_add(1)
            .saturating_add(1)
            .saturating_add(1)],
        n => [n],
    })
    .flatten();

Works just fine in stable, so it's likely #8520

[Serial-ATA]

Ah, just read into it. Looks like the real issue is rust-lang/rustfix#141. ...or is it? I'll have to look into it more.

[Serial-ATA]

This is actually a Cargo issue. rust-lang/rustfix@635eef7 didn't make it into rustfix 0.6.0, which Cargo depends on. Just have to wait for a new release.

[droogmic]

I ran into this as well, I see we are waiting for an answer on rust-lang/rustfix#211.

[hellow554]

The culprit should be solved by updating rustfix to 0.6.1, but @Serial-ATA discovered a very interesting case:

fn issue8734() {
    let _ = [0u8, 1, 2, 3]
        .into_iter()
        .map(|n| match n {
            1 => [n
                .saturating_add(1)
                .saturating_add(1)
                .saturating_add(1)
                .saturating_add(1)
                .saturating_add(1)
                .saturating_add(1)
                .saturating_add(1)
                .saturating_add(1)],
            n => [n],
        })
        .flatten();
}

gets autocorrect to

fn issue8734() {
    let _ = [0u8, 1, 2, 3]
        .into_iter()
        .flat_map(|n| match n {
            1 => [n
                .saturating_add(1)
            1 => [n
                .saturating_add(1)
                .saturating_add(1)
                .saturating_add(1)
                .saturating_add(1)
                .saturating_add(1)
                .saturating_add(1)
                .saturating_add(1)
                .saturating_add(1)],
            n => [n],
        });
}

which is weird and incorrect.

Let's investigate this further