ICE: "Tried to access field 0 of union with 0 fields"

[bkchr]

I get the following internal compiler error with rustc nightly:

thread 'rustc' panicked at 'Tried to access field 0 of union with 0 fields', src/librustc_target/abi/mod.rs:742:17
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace.

error: internal compiler error: unexpected panic

note: the compiler unexpectedly panicked. this is a bug.

note: we would appreciate a bug report: https:/rust-lang/rust/blob/master/CONTRIBUTING.md#bug-reports

note: rustc 1.40.0-nightly (e413dc36a 2019-10-14) running on x86_64-unknown-linux-gnu

note: compiler flags: -C debuginfo=2 -C incremental --crate-type lib

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

To reproduce, just checkout the repo at the following commit: paritytech/polkadot@3174117

Run: BUILD_DUMMY_WASM_BINARY=1 cargo build -p polkadot-service

[Centril]

To reproduce, just checkout the repo at the following commit: paritytech/polkadot@3174117

Can you please provide a minimized example that would fit in the playground without any dependencies? We would also appreciate a backtrace.

[bkchr]

Sorry for forgetting the backtrace, here we go:

thread 'rustc' panicked at 'Tried to access field 0 of union with 0 fields', src/librustc_target/abi/mod.rs:742:17
stack backtrace:
   0: backtrace::backtrace::libunwind::trace
             at /cargo/registry/src/github.com-1ecc6299db9ec823/backtrace-0.3.37/src/backtrace/libunwind.rs:88
   1: backtrace::backtrace::trace_unsynchronized
             at /cargo/registry/src/github.com-1ecc6299db9ec823/backtrace-0.3.37/src/backtrace/mod.rs:66
   2: std::sys_common::backtrace::_print_fmt
             at src/libstd/sys_common/backtrace.rs:77
   3: <std::sys_common::backtrace::_print::DisplayBacktrace as core::fmt::Display>::fmt
             at src/libstd/sys_common/backtrace.rs:61
   4: core::fmt::write
             at src/libcore/fmt/mod.rs:1028
   5: std::io::Write::write_fmt
             at src/libstd/io/mod.rs:1412
   6: std::sys_common::backtrace::_print
             at src/libstd/sys_common/backtrace.rs:65
   7: std::sys_common::backtrace::print
             at src/libstd/sys_common/backtrace.rs:50
   8: std::panicking::default_hook::{{closure}}
             at src/libstd/panicking.rs:189
   9: std::panicking::default_hook
             at src/libstd/panicking.rs:206
  10: rustc_driver::report_ice
  11: std::panicking::rust_panic_with_hook
             at src/libstd/panicking.rs:473
  12: std::panicking::continue_panic_fmt
             at src/libstd/panicking.rs:376
  13: std::panicking::begin_panic_fmt
             at src/libstd/panicking.rs:331
  14: rustc_target::abi::FieldPlacement::offset
  15: rustc_codegen_ssa::mir::place::PlaceRef<V>::project_field
  16: rustc_codegen_ssa::mir::place::<impl rustc_codegen_ssa::mir::FunctionCx<Bx>>::codegen_place
  17: rustc_codegen_ssa::mir::codegen_mir
  18: rustc_codegen_ssa::base::codegen_instance
  19: <rustc::mir::mono::MonoItem as rustc_codegen_ssa::mono_item::MonoItemExt>::define
  20: rustc_codegen_llvm::base::compile_codegen_unit::module_codegen
  21: rustc::dep_graph::graph::DepGraph::with_task
  22: rustc_codegen_llvm::base::compile_codegen_unit
  23: rustc_codegen_ssa::base::codegen_crate
  24: <rustc_codegen_llvm::LlvmCodegenBackend as rustc_codegen_utils::codegen_backend::CodegenBackend>::codegen_crate
  25: rustc_interface::passes::start_codegen::{{closure}}
  26: rustc::util::common::time
  27: rustc_interface::passes::BoxedGlobalCtxt::access::{{closure}}
  28: rustc_interface::passes::create_global_ctxt::{{closure}}
  29: rustc_interface::queries::Query<T>::compute
  30: rustc_interface::queries::<impl rustc_interface::interface::Compiler>::ongoing_codegen
  31: rustc_interface::interface::run_compiler_in_existing_thread_pool
  32: std::thread::local::LocalKey<T>::with
  33: scoped_tls::ScopedKey<T>::set
  34: syntax::with_globals

Regarding a minimized example, I would really like to provide you one, but I did not change any code. I just updated the nightly compiler. The latest stable compiles this code without any problems.

[Centril]

Regarding a minimized example, I would really like to provide you one, but I did not change any code. I just updated the nightly compiler. The latest stable compiles this code without any problems.

Sure; it's just substantially more difficult to fix a bug when the input is large as compared to just a few lines of code you can put in a file (and in the playground!). :) Reports with a minimal snippet tend to get fixed much faster and we sorta need a regression test as well.

[Centril]

cc @eddyb @nagisa

[Centril]

The latest stable compiles this code without any problems.

Did you try the beta compiler? Does it work there?

[bkchr]

Okay, I tried to find the problem and to reproduce it. I failed at reproducing :(

However, I found the line https:/paritytech/polkadot/blob/master/service/src/lib.rs#L167
If I remove this call to build() it does not crash.

This build call is implemented in the following monstrous impl block: https:/paritytech/substrate/blob/master/core/service/src/builder.rs#L803-L843

I somehow suspect that the where clause could be the reason for it.

Sorry for not being able to reduce it anymore!

I'm currently testing the beta channel and will report back when I know more.

[bkchr]

Beta works for me.

[bkchr]

I downgraded to "rustc 1.40.0-nightly (c27f756 2019-10-13)" and that compiles the crate without any problems.

[pnkfelix]

triage: P-high. self-assigning. removing nomination tag.

[koute]

Minimal example:

enum Empty {}
enum Enum {
    Empty( Empty )
}

fn foobar() -> Option< Enum > {
    let value: Option< Empty > = None;
    Some( Enum::Empty( value? ) )
}

fn main() {
    foobar();
}

Playground link.

It works on stable and on beta; it ICEs on nightly.

[Centril]

Curiously this does not reproduce with ! instead of Empty.

cc @nagisa @eddyb @rkruppe

[eddyb]

I believe this is an assert @oli-obk added at my request - it already existed in miri, but was missing from the original implementation in rustc_target::abi.

[Demi-Marie]

@eddyb can we remove the assertion for now?

[eddyb]

Oops, this slipped into beta (@oli-obk was on vacation and I forgot about it completely).

[pnkfelix]

(to be fair, it was assigned to me, and I just kept letting it slip on my own priority list as well...)

[pnkfelix]

reassigning to @oli-obk based on feedback from compiler team

[spastorino]

This regressed here