ICE when using unboxed closures with lifetime HRTBs

[abonander]

Most basic test case:

#![feature(unboxed_closures)]

#[deriving(Show)]
struct LifetimeStruct<'a>;

fn main() {
    takes_hrtb_closure(|&mut: lts| println!("{}", lts));
}

fn takes_hrtb_closure<F: for<'a>FnMut(LifetimeStruct<'a>)>(f: F) {
    f(LifetimeStruct);
}

Compiler Output:

$ RUST_BACKTRACE=1 rustc hrtb_test.rs 
error: internal compiler error: unexpected panic
note: the compiler unexpectedly panicked. this is a bug.
note: we would appreciate a bug report: http://doc.rust-lang.org/complement-bugreport.html
note: run with `RUST_BACKTRACE=1` for a backtrace
task 'rustc' panicked at 'assertion failed: !ty::type_needs_infer(ty)', /home/rustbuild/src/rust-buildbot/slave/nightly-linux/build/src/librustc/middle/typeck/mod.rs:290

stack backtrace:
   1:     0x7fde72dc4a80 - rt::backtrace::imp::write::had71017ae03f24a0vVs
   2:     0x7fde72dc7b10 - failure::on_fail::he71b2ccb43241291iht
   3:     0x7fde7358c0e0 - unwind::begin_unwind_inner::he7edbf4992e36292Pad
   4:     0x7fde71b53790 - unwind::begin_unwind::h16417137682926516883
   5:     0x7fde71ecab50 - middle::typeck::write_ty_to_tcx::h1c217af62028a95brtK
   6:     0x7fde71ec62e0 - middle::typeck::check::writeback::WritebackCx<'cx, 'tcx>::visit_node_id::h0cabe78aea9d5af18n9
   7:     0x7fde71ec4410 - middle::typeck::check::writeback::WritebackCx<'cx, 'tcx>.Visitor<'v>::visit_expr::he3eee31f7078a8284c9
   8:     0x7fde71ec4410 - middle::typeck::check::writeback::WritebackCx<'cx, 'tcx>.Visitor<'v>::visit_expr::he3eee31f7078a8284c9
   9:     0x7fde71ec9e40 - visit::walk_block::h11911811437128181636
  10:     0x7fde71ec6050 - middle::typeck::check::writeback::resolve_type_vars_in_fn::h1f622545b4b2c27c978
  11:     0x7fde71f25df0 - middle::typeck::check::check_bare_fn::hd3d74fd6bc764cbdhfh
  12:     0x7fde71f21e60 - middle::typeck::check::check_item::h40f208d380d8ba92izh
  13:     0x7fde71f25bb0 - middle::typeck::check::check_item_types::hab579f21022c1d0freh
  14:     0x7fde722ff7e0 - util::common::time::h2047692826153001907
  15:     0x7fde722fe960 - middle::typeck::check_crate::h2edb3a53d123f8c0eMK
  16:     0x7fde73a81500 - driver::driver::phase_3_run_analysis_passes::ha5acf8a70ce48fbfMTR
  17:     0x7fde73a76320 - driver::driver::compile_input::h2b7fa47cad7323abwAR
  18:     0x7fde73af69f0 - driver::run_compiler::hd1dcd9c26f83ca38HyT
  19:     0x7fde73af68e0 - driver::run::closure.53542
  20:     0x7fde73913400 - task::TaskBuilder<S>::try_future::closure.38917
  21:     0x7fde739131f0 - task::TaskBuilder<S>::spawn_internal::closure.38888
  22:     0x7fde73e12b50 - task::NativeSpawner.Spawner::spawn::closure.2470
  23:     0x7fde735e4b40 - rust_try_inner
  24:     0x7fde735e4b30 - rust_try
  25:     0x7fde73589a40 - unwind::try::h074ea65506c07a5arZc
  26:     0x7fde735898d0 - task::Task::run::h7e964b829504fb5ah5b
  27:     0x7fde73e12890 - task::NativeSpawner.Spawner::spawn::closure.2396
  28:     0x7fde7358b0f0 - thread::thread_start::hd81953be3de79f59nqc
  29:     0x7fde6deab0c0 - start_thread
  30:     0x7fde73250f89 - __clone
  31:                0x0 - <unknown>

This is a boiled-down version of the problem I'm having. It's not giving them here, but in the original program where I discovered this, I did get compiler lifetime errors:

/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:94: 19:101 error: cannot infer an appropriate lifetime for borrow expression due to conflicting requirements
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19         multipart.foreach_entry(|&mut: name, content| println!("Name: {} Content: {}", name, content));
                                                                                                                                                        ^~~~~~~
note: in expansion of format_args!
<std macros>:2:23: 2:77 note: expansion site
<std macros>:1:1: 3:2 note: in expansion of println!
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:55: 19:103 note: expansion site
note: first, the lifetime cannot outlive lifetime ReInfer(ReSkolemized(10, BrNamed(DefId { krate: 30, node: 1704 }, "'a"(2090))))...
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:94: 19:101 note: ...so that the reference type `&multipart::MultipartField<'a>` does not outlive the data it points at
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19         multipart.foreach_entry(|&mut: name, content| println!("Name: {} Content: {}", name, content));
                                                                                                                                                        ^~~~~~~
note: in expansion of format_args!
<std macros>:2:23: 2:77 note: expansion site
<std macros>:1:1: 3:2 note: in expansion of println!
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:55: 19:103 note: expansion site
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:94: 19:101 note: but, the lifetime must be valid for the expression at 19:93...
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19         multipart.foreach_entry(|&mut: name, content| println!("Name: {} Content: {}", name, content));
                                                                                                                                                        ^~~~~~~
note: in expansion of format_args!
<std macros>:2:23: 2:77 note: expansion site
<std macros>:1:1: 3:2 note: in expansion of println!
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:55: 19:103 note: expansion site
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:94: 19:101 note: ...so that reference is valid at the time of borrow
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19         multipart.foreach_entry(|&mut: name, content| println!("Name: {} Content: {}", name, content));
                                                                                                                                                        ^~~~~~~
note: in expansion of format_args!
<std macros>:2:23: 2:77 note: expansion site
<std macros>:1:1: 3:2 note: in expansion of println!
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:55: 19:103 note: expansion site
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:19: 19:103 error: cannot infer an appropriate lifetime due to conflicting requirements
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19         multipart.foreach_entry(|&mut: name, content| println!("Name: {} Content: {}", name, content));
                                                                             ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
note: first, the lifetime cannot outlive lifetime ReInfer(ReSkolemized(10, BrNamed(DefId { krate: 30, node: 1704 }, "'a"(2090))))...
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:19: 19:103 note: ...so that trait type parameters matches those specified on the impl (expected `for<'a> core::ops::FnMut<(collections::string::String, multipart::MultipartField<'a>), ()>`, found `core::ops::FnMut<(collections::string::String, multipart::MultipartField<'a>), ()>`)
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19         multipart.foreach_entry(|&mut: name, content| println!("Name: {} Content: {}", name, content));
                                                                             ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:94: 19:101 note: but, the lifetime must be valid for the expression at 19:93...
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19         multipart.foreach_entry(|&mut: name, content| println!("Name: {} Content: {}", name, content));
                                                                                                                                                        ^~~~~~~
note: in expansion of format_args!
<std macros>:2:23: 2:77 note: expansion site
<std macros>:1:1: 3:2 note: in expansion of println!
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:55: 19:103 note: expansion site
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:94: 19:101 note: ...so that type parameter instantiated with `multipart::MultipartField<'_>`, will meet its declared lifetime bounds.
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19         multipart.foreach_entry(|&mut: name, content| println!("Name: {} Content: {}", name, content));
                                                                                                                                                        ^~~~~~~
note: in expansion of format_args!
<std macros>:2:23: 2:77 note: expansion site
<std macros>:1:1: 3:2 note: in expansion of println!
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:55: 19:103 note: expansion site
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:94: 19:101 error: type of expression contains references that are not valid during the expression: `multipart::MultipartField<'a>`
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19         multipart.foreach_entry(|&mut: name, content| println!("Name: {} Content: {}", name, content));
                                                                                                                                                        ^~~~~~~
note: in expansion of format_args!
<std macros>:2:23: 2:77 note: expansion site
<std macros>:1:1: 3:2 note: in expansion of println!
/home/austin/Rust/multipart/src/bin/multipart_server.rs:19:55: 19:103 note: expansion site
note: type is only valid for lifetime ReInfer(ReSkolemized(10, BrNamed(DefId { krate: 30, node: 1704 }, "'a"(2090))))
error: internal compiler error: unexpected panic
note: the compiler unexpectedly panicked. this is a bug.
note: we would appreciate a bug report: http://doc.rust-lang.org/complement-bugreport.html
note: run with `RUST_BACKTRACE=1` for a backtrace
task 'rustc' panicked at 'assertion failed: !ty::type_needs_infer(ty)', /home/rustbuild/src/rust-buildbot/slave/nightly-linux/build/src/librustc/middle/typeck/mod.rs:290

(Backtrace omitted because it was the same.)

ICE Location
multipart.foreach_entry() definition

This seems like a personally reasonable use case, and I believe my logic here is sound. It should work just fine with a static function or an old boxed closure, but I'd really like to make it work with unboxed closures.

[nikomatsakis]

cc me

[nikomatsakis]

Hmm. Seems to work for me, rebuilding with vanilla master to double check.

[nikomatsakis]

Yes, works for me.

[nikomatsakis]

Marking as needs-test