ICE: enum discriminant depends on generic arguments

[Patryk27]

Following code ICEs on today's nightly (97e58c0 2019-09-20):

enum Foo {
  A = foo(),
  B = foo(), // <- error: internal compiler error: src/librustc/ty/mod.rs:2400: enum discriminant depends on generic arguments
}

const fn foo<T>() -> isize {
   0
}

What's peculiar is that no ICE is shown when the Foo enum contains only one item:

enum Foo {
  A = foo(),
}

[Centril]

Backtrace:

error: internal compiler error: src/librustc/ty/mod.rs:2400: enum discriminant depends on generic arguments
 --> src/lib.rs:3:7
  |
3 |   B = foo(),
  |       ^^^^^

thread 'rustc' panicked at 'Box<Any>', src/librustc_errors/lib.rs:578:9
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:76
   3: <std::sys_common::backtrace::_print::DisplayBacktrace as core::fmt::Display>::fmt
             at src/libstd/sys_common/backtrace.rs:60
   4: core::fmt::write
             at src/libcore/fmt/mod.rs:1030
   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:64
   7: std::sys_common::backtrace::print
             at src/libstd/sys_common/backtrace.rs:49
   8: std::panicking::default_hook::{{closure}}
             at src/libstd/panicking.rs:196
   9: std::panicking::default_hook
             at src/libstd/panicking.rs:210
  10: rustc_driver::report_ice
  11: std::panicking::rust_panic_with_hook
             at src/libstd/panicking.rs:477
  12: std::panicking::begin_panic
  13: rustc_errors::Handler::span_bug
  14: rustc::util::bug::opt_span_bug_fmt::{{closure}}
  15: rustc::ty::context::tls::with_opt::{{closure}}
  16: rustc::ty::context::tls::with_context_opt
  17: rustc::ty::context::tls::with_opt
  18: rustc::util::bug::opt_span_bug_fmt
  19: rustc::util::bug::span_bug_fmt
  20: rustc::ty::AdtDef::eval_explicit_discr
  21: <rustc_typeck::collect::CollectItemTypesVisitor as rustc::hir::intravisit::Visitor>::visit_item
  22: rustc::hir::map::Map::visit_item_likes_in_module
  23: rustc_typeck::collect::collect_mod_item_types
  24: rustc::ty::query::__query_compute::collect_mod_item_types
  25: rustc::ty::query::<impl rustc::ty::query::config::QueryAccessors for rustc::ty::query::queries::collect_mod_item_types>::compute
  26: rustc::dep_graph::graph::DepGraph::with_task_impl
  27: rustc::ty::query::plumbing::<impl rustc::ty::context::TyCtxt>::get_query
  28: rustc_typeck::check_crate::{{closure}}::{{closure}}
  29: rustc::util::common::time
  30: rustc_typeck::check_crate
  31: rustc_interface::passes::analysis
  32: rustc::ty::query::__query_compute::analysis
  33: rustc::ty::query::plumbing::<impl rustc::ty::context::TyCtxt>::get_query
  34: rustc_interface::passes::BoxedGlobalCtxt::access::{{closure}}
  35: rustc_interface::passes::create_global_ctxt::{{closure}}
  36: rustc_interface::interface::run_compiler_in_existing_thread_pool
  37: std::thread::local::LocalKey<T>::with
  38: scoped_tls::ScopedKey<T>::set
  39: syntax::with_globals
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

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

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

note: rustc 1.39.0-nightly (97e58c0d3 2019-09-20) running on x86_64-unknown-linux-gnu

note: compiler flags: -C codegen-units=1 -C debuginfo=2 --crate-type lib

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

query stack during panic:
#0 [collect_mod_item_types] collecting item types in top-level module
#1 [analysis] running analysis passes on this crate
end of query stack
error: aborting due to 3 previous errors

Some errors have detailed explanations: E0019, E0282.
For more information about an error, try `rustc --explain E0019`.
error: could not compile `playground`.

[varkor]

What's peculiar is that no ICE is shown when the Foo enum contains only one item:

This is because the discriminant assignment code tries to increment the discriminant of the previous variant, which was an error, and thus fails in a different way to the first. However, the code looks like it handles that case, so I'm not yet sure what's going wrong here.

Edit: I think there's some state changing inside const_eval — the discriminant assignment code seems to be working as expected. const_eval is being called with essentially identical arguments, but something about the first error is causing the second one to behave differently.

[oli-obk]

It's not a regression. Stable shows

   Compiling playground v0.0.1 (/playground)
error[E0282]: type annotations needed
 --> src/lib.rs:2:7
  |
2 |   A = foo(),
  |       ^^^ cannot infer type for `T`

error[E0019]: constant contains unimplemented expression type
 --> src/lib.rs:3:7
  |
3 |   B = foo(), // <- error: internal compiler error: src/librustc/ty/mod.rs:2400: enum discriminant depends on generic arguments
  |       ^^^^^

error: internal compiler error: src/librustc/ty/mod.rs:2394: enum discriminant depends on generic arguments
 --> src/lib.rs:3:7
  |
3 |   B = foo(), // <- error: internal compiler error: src/librustc/ty/mod.rs:2400: enum discriminant depends on generic arguments
  |    

I don't see how we could be keeping state. Maybe it's happening because there's no "cannot infer type for T" error on anything but the first item? So A is never evaluated, thus not causing an ICE.

[varkor]

Maybe it's happening because there's no "cannot infer type for T" error on anything but the first item?

Hmm, perhaps. In that case, it's simplest just to delay_span_bug instead.

[varkor]

This no longer ICEs on nightly, though it does ICE on beta. We now get type inference errors on both variants.

[varkor]

Reopening, as we should have a test.