Existental type, const fn, and generator ICE

[DutchGhost]

The following constant function that returns a Generator, ICE's:

#![feature(const_fn)]
#![feature(generators)]
#![feature(generator_trait)]
#![feature(existential_type)]

use std::ops::Generator;
 
existential type GenOnce<Y, R>: Generator<Yield = Y, Return = R>;

const fn const_generator<Y, R>(yielding: Y, returning: R) -> GenOnce<Y, R> {
    move || {
        yield yielding;
        
        return returning;
    }
}

const foo: GenOnce<usize, usize> = const_generator(10, 100);

fn main() {
    
}

Backtrace:

thread 'main' panicked at 'index out of bounds: the len is 2 but the index is 2', /checkout/src/librustc_mir/interpret/eval_context.rs:1851:68
stack backtrace:
   0: std::sys::unix::backtrace::tracing::imp::unwind_backtrace
             at libstd/sys/unix/backtrace/tracing/gcc_s.rs:49
   1: std::sys_common::backtrace::print
             at libstd/sys_common/backtrace.rs:71
             at libstd/sys_common/backtrace.rs:59
   2: std::panicking::default_hook::{{closure}}
             at libstd/panicking.rs:211
   3: std::panicking::default_hook
             at libstd/panicking.rs:227
   4: rustc::util::common::panic_hook
   5: std::panicking::rust_panic_with_hook
             at libstd/panicking.rs:479
   6: std::panicking::continue_panic_fmt
             at libstd/panicking.rs:390
   7: rust_begin_unwind
             at libstd/panicking.rs:325
   8: core::panicking::panic_fmt
             at libcore/panicking.rs:77
   9: core::panicking::panic_bounds_check
             at libcore/panicking.rs:59
  10: rustc::ty::<impl rustc::ty::context::TyCtxt<'a, 'gcx, 'tcx>>::with_freevars
  11: <rustc_mir::interpret::eval_context::EvalContext<'a, 'mir, 'tcx, M>>::write_field_name
  12: <rustc_mir::interpret::eval_context::EvalContext<'a, 'mir, 'tcx, M>>::validate_ptr_target
  13: rustc_lint::builtin::check_const
  14: <rustc_lint::register_builtins::BuiltinCombinedLateLintPass as rustc::lint::LateLintPass<'a, 'tcx>>::check_item
  15: <rustc::lint::context::LateContext<'a, 'tcx> as rustc::hir::intravisit::Visitor<'tcx>>::visit_item
  16: <rustc::lint::context::LateContext<'a, 'tcx> as rustc::hir::intravisit::Visitor<'tcx>>::visit_mod
  17: rustc::hir::intravisit::walk_crate
  18: rustc::lint::context::check_crate
  19: rustc::util::common::time
  20: rustc::ty::context::tls::enter_context
  21: <std::thread::local::LocalKey<T>>::with
  22: rustc::ty::context::TyCtxt::create_and_enter
  23: rustc_driver::driver::compile_input
  24: rustc_driver::run_compiler_with_pool
  25: syntax::with_globals
  26: <std::panic::AssertUnwindSafe<F> as core::ops::function::FnOnce<()>>::call_once
  27: __rust_maybe_catch_panic
             at libpanic_unwind/lib.rs:102
  28: rustc_driver::run
  29: rustc_driver::main
  30: std::rt::lang_start::{{closure}}
  31: std::panicking::try::do_call
             at libstd/rt.rs:59
             at libstd/panicking.rs:310
  32: __rust_maybe_catch_panic
             at libpanic_unwind/lib.rs:102
  33: std::rt::lang_start_internal
             at libstd/panicking.rs:289
             at libstd/panic.rs:392
             at libstd/rt.rs:58
  34: main
  35: __libc_start_main
  36: <unknown>
query stack during panic:
end of query stack

[DutchGhost]

This no longer ICE's!

[oli-obk]

New ICE:

error: internal compiler error: broken MIR in DefId(0/0:6 ~ playground[d6cf]::foo[0]) (bb0[0]): equate_inputs_and_outputs: `GenOnce::<usize, usize>==GenOnce::<usize, usize>` failed with `NoSolution`
  --> src/main.rs:18:1
   |
18 | const foo: GenOnce<usize, usize> = const_generator(10, 100);
   | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

[Aaron1011]

This no longer ICEs on the latest nightly.

[oli-obk]

I'm assuming this (and #54899, #60564, #60407) were fixed by #62423 ?

If so, we could close all four by adding their failing examples as regression tests (if there aren't already such tests).

[Centril]

If so, we could close all four by adding their failing examples as regression tests (if there aren't already such tests).

I would prefer to have redundant tests here rather than under-testing so some bit of duplication is fine imo.

Also cc @varkor @alexreg -- we should add these tests before we change the syntax ideally.