#[derive(Debug)] on #[repr(packed)] enum causes internal compiler error

[urinacon]

Code

#[derive(Debug)]
#[repr(packed)]
enum COption<T> {
    None,
    Some(T),
}

fn main() {
}

Meta

rustc --version --verbose:

rustc 1.82.0 (f6e511eec 2024-10-15) (Fedora 1.82.0-1.fc40)
binary: rustc
commit-hash: f6e511eec7342f59a25f7c0534f1dbea00d01b14
commit-date: 2024-10-15
host: x86_64-unknown-linux-gnu
release: 1.82.0
LLVM version: 18.1.8

Error output

error: internal compiler error: compiler/rustc_mir_transform/src/check_packed_ref.rs:49:21: builtin derive created an unaligned reference
 --> src/main.rs:5:10
  |
1 | #[derive(Debug)]
  |          ----- in this derive macro expansion
...
5 |     Some(T),
  |          ^
  |
  = note: this error: internal compiler error originates in the derive macro `Debug` (in Nightly builds, run with -Z macro-backtrace for more info)

thread 'rustc' panicked at compiler/rustc_mir_transform/src/check_packed_ref.rs:49:21:
Box<dyn Any>
stack backtrace:
   0:     0x7fd20cb85b78 - <std::sys::backtrace::BacktraceLock::print::DisplayBacktrace as core::fmt::Display>::fmt::h1ceeeb1b55e290c2
   1:     0x7fd20c582b9b - core::fmt::write::hb3a48913f78b0dae
   2:     0x7fd20cb78ba3 - <unknown>
   3:     0x7fd20cb887ea - <unknown>
   4:     0x7fd20cb8842c - std::panicking::default_hook::h49089136d7ad7532
   5:     0x7fd20a1425ad - <unknown>
   6:     0x7fd20cb891e6 - std::panicking::rust_panic_with_hook::ha43241025fb228f0
   7:     0x7fd20a18c221 - <unknown>
   8:     0x7fd20a17c246 - <unknown>
   9:     0x7fd20a17bfc6 - <unknown>
  10:     0x7fd20a19a8e1 - <rustc_errors[9a66caa3f8a65bb2]::diagnostic::BugAbort as rustc_errors[9a66caa3f8a65bb2]::diagnostic::EmissionGuarantee>::emit_producing_guarantee
  11:     0x7fd20a92c36d - <unknown>
  12:     0x7fd20a96b188 - <unknown>
  13:     0x7fd20a96bada - <unknown>
  14:     0x7fd20a9570db - <unknown>
  15:     0x7fd20a94dd57 - <unknown>
  16:     0x7fd20be2d539 - <unknown>
  17:     0x7fd20bf150eb - rustc_mir_transform[ec94f2ac466ce048]::mir_built
  18:     0x7fd20c1c6467 - <unknown>
  19:     0x7fd20ac4e640 - <unknown>
  20:     0x7fd20ac3cb28 - <unknown>
  21:     0x7fd20ad4ed98 - <unknown>
  22:     0x7fd20bd63e17 - rustc_mir_build[b87ccf4b6ee18353]::check_unsafety::check_unsafety
  23:     0x7fd20c1b87f7 - <unknown>
  24:     0x7fd20ac4dd90 - <unknown>
  25:     0x7fd20ac3714e - <unknown>
  26:     0x7fd20ad5d45f - <unknown>
  27:     0x7fd20baa8cbb - <unknown>
  28:     0x7fd20c8bd509 - rustc_interface[73b4cfe22fadb0e6]::passes::analysis
  29:     0x7fd20ca728b7 - <unknown>
  30:     0x7fd20ac4d70f - <unknown>
  31:     0x7fd20ac06129 - <unknown>
  32:     0x7fd20ad49194 - <unknown>
  33:     0x7fd20c7b0ee4 - <unknown>
  34:     0x7fd20c7ad945 - <unknown>
  35:     0x7fd20c7b3dd0 - <unknown>
  36:     0x7fd20cb952ab - <unknown>
  37:     0x7fd2094a66d7 - start_thread
  38:     0x7fd20952a60c - __clone3
  39:                0x0 - <unknown>

note: we would appreciate a bug report: https://github.com/rust-lang/rust/issues/new?labels=C-bug%2C+I-ICE%2C+T-compiler&template=ice.md

note: rustc 1.82.0 (f6e511eec 2024-10-15) (Fedora 1.82.0-1.fc40) running on x86_64-unknown-linux-gnu

note: compiler flags: --crate-type bin -C embed-bitcode=no -C debuginfo=2 -C incremental=[REDACTED]

note: some of the compiler flags provided by cargo are hidden

query stack during panic:
#0 [mir_built] building MIR for `<impl at src/main.rs:1:10: 1:15>::fmt`
#1 [check_unsafety] unsafety-checking `<impl at src/main.rs:1:10: 1:15>::fmt`
end of query stack

Backtrace

error: internal compiler error: compiler/rustc_mir_transform/src/check_packed_ref.rs:49:21: builtin derive created an unaligned reference
 --> src/main.rs:5:10
  |
1 | #[derive(Debug)]
  |          ----- in this derive macro expansion
...
5 |     Some(T),
  |          ^
  |
  = note: this error: internal compiler error originates in the derive macro `Debug` (in Nightly builds, run with -Z macro-backtrace for more info)

thread 'rustc' panicked at compiler/rustc_mir_transform/src/check_packed_ref.rs:49:21:
Box<dyn Any>
stack backtrace:
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

note: we would appreciate a bug report: https://github.com/rust-lang/rust/issues/new?labels=C-bug%2C+I-ICE%2C+T-compiler&template=ice.md

note: rustc 1.82.0 (f6e511eec 2024-10-15) (Fedora 1.82.0-1.fc40) running on x86_64-unknown-linux-gnu

note: compiler flags: --crate-type bin -C embed-bitcode=no -C debuginfo=2 -C incremental=[REDACTED]

note: some of the compiler flags provided by cargo are hidden

query stack during panic:
#0 [mir_built] building MIR for `<impl at src/main.rs:1:10: 1:15>::fmt`
#1 [check_unsafety] unsafety-checking `<impl at src/main.rs:1:10: 1:15>::fmt`
#2 [analysis] running analysis passes on this crate
end of query stack

[Noratrieb]

error[E0517]: attribute should be applied to a struct or union
 --> src/main.rs:2:8
  |
2 |   #[repr(packed)]
  |          ^^^^^^
3 | / enum COption<T> {
4 | |     None,
5 | |     Some(T),
6 | | }
  | |_- not a struct or union

For more information about this error, try `rustc --explain E0517`.

repr(packed) on enums is not allowed. I suspect the easiest fix is to make the derive macro error or expand to nothing when it sees repr(packed). alternatively, the MIR transform could just ignore enums.

[zachs18]

This appears to have regressed in 1.78.0 on stable (1.77.0 does not ICE), or nightly-2024-02-09.

bisection

Bisection determined this regressed in nightly-2024-02-09, but there's no CI artifacts from that far back, so I'm not sure which PR led to the ICE.

Bisected with cargo bisect-rustc --start=1.77.0 --end=1.78.0 --regress ice

searched toolchains nightly-2024-02-01 through nightly-2024-03-16


********************************************************************************
Regression in nightly-2024-02-09
********************************************************************************

fetching https://static.rust-lang.org/dist/2024-02-08/channel-rust-nightly-git-commit-hash.txt
nightly manifest 2024-02-08: 40 B / 40 B [=====================================================================] 100.00 % 846.46 KB/s converted 2024-02-08 to 8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6
fetching https://static.rust-lang.org/dist/2024-02-09/channel-rust-nightly-git-commit-hash.txt
nightly manifest 2024-02-09: 40 B / 40 B [=====================================================================] 100.00 % 521.58 KB/s converted 2024-02-09 to 98aa3624be70462d6a25ed5544333e3df62f4c66
looking for regression commit between 2024-02-08 and 2024-02-09
fetching (via remote github) commits from max(8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6, 2024-02-06) to 98aa3624be70462d6a25ed5544333e3df62f4c66
ending github query because we found starting sha: 8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6
get_commits_between returning commits, len: 9
  commit[0] 2024-02-07: Auto merge of #120748 - Nadrieril:rollup-dj0qwv5, r=Nadrieril
  commit[1] 2024-02-08: Auto merge of #120381 - fee1-dead-contrib:reconstify-add, r=compiler-errors
  commit[2] 2024-02-08: Auto merge of #120521 - reitermarkus:generic-nonzero-constructors, r=dtolnay
  commit[3] 2024-02-08: Auto merge of #120558 - oli-obk:missing_impl_item_ice, r=estebank
  commit[4] 2024-02-08: Auto merge of #120579 - GuillaumeGomez:prevent-running-unneeded-code, r=notriddle
  commit[5] 2024-02-08: Auto merge of #120550 - oli-obk:track_errors8, r=estebank
  commit[6] 2024-02-08: Auto merge of #120767 - matthiaskrgr:rollup-0k8ib1c, r=matthiaskrgr
  commit[7] 2024-02-08: Auto merge of #120544 - BoxyUwU:enter_forall, r=lcnr
  commit[8] 2024-02-08: Auto merge of #120807 - matthiaskrgr:rollup-1pf3glu, r=matthiaskrgr
ERROR: no CI builds available between 8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6 and 98aa3624be70462d6a25ed5544333e3df62f4c66 within last 167 days

[matthiaskrgr]

Duplicate of #120873 I think

[Noratrieb]

don't think so

[matthiaskrgr]

What difference do you see?
The stacktrace of

#[derive(Debug)]
#[repr(packed)]
enum COption<T> {
    None,
    Some(T),
}

fn main() {
}

is identical to #120873 (comment)

#[repr(packed)]

struct Dealigned<T>(u8, T);

#[derive(PartialEq)]
#[repr(C)]
struct Dealigned<T>(u8, T);

[Noratrieb]

Yes, it's the exact same check firing. But I think the check is firing for entirely different reasons. It could be the same but I'd be surprised. It's nice to have the links between the two issues because they are similar but I certainly wouldn't close this.

[el-ev]

Introduced by eab2adb.

[moxian]

Both this and #120873 bisect same way to nightly-2024-02-09
Even though the ICE has been triggerable since 2017 (#120873 (comment))

Either way,
@rustbot label: -E-needs-bisection