ICE on resolve error inside impl block

[alexcrichton]

trait Foo {}        
impl Foo for *int { 
    fn foo() {      
        try_finally 
    }               
}                  

error: internal compiler error: unexpected failure
note: the compiler hit an unexpected failure path. this is a bug.
note: we would appreciate a bug report: http://static.rust-lang.org/doc/master/complement-bugreport.html
note: run with `RUST_BACKTRACE=1` for a backtrace
task 'rustc' failed at 'internal error: entered unreachable code', /Users/rustbuild/src/rust-buildbot/slave/nightly-mac/build/src/librustc/middle/resolve.rs:4937
stack backtrace:
   1:        0x103faac8b - rt::backtrace::imp::write::h6bf09e1ae0cfa18aoBz::v0.11.0.pre
   2:        0x103f2deb1 - rt::unwind::begin_unwind_inner::he3bc737d2c39ab1cwbz::v0.11.0.pre
   3:        0x101127aed - rt::unwind::begin_unwind::h7336716533461090442::v0.11.0.pre
   4:        0x10130c086 - middle::resolve::Resolver<'a>::resolve_expr::hfb7d6b3fa4ebd679nqY::v0.11.0.pre
   5:        0x101363fe0 - visit::walk_expr_opt::h4994540791083417155::v0.11.0.pre
   6:        0x101309a07 - middle::resolve::Resolver<'a>::resolve_block::hc32272e12cab421fP6W::v0.11.0.pre
   7:        0x10135a4f0 - middle::resolve::Resolver<'a>::resolve_function::closure.67175
   8:        0x10132e4a2 - middle::resolve::Resolver<'a>::with_type_parameter_rib::h9764680d6cf0a5cfEsW::v0.11.0.pre
   9:        0x101358981 - middle::resolve::Resolver<'a>::resolve_function::he939ff208f5c8b222xW::v0.11.0.pre
  10:        0x101361648 - middle::resolve::Resolver<'a>::resolve_implementation::closure.67235
  11:        0x10136050b - middle::resolve::Resolver<'a>::with_current_self_type::h3499330292536224027::v0.11.0.pre
  12:        0x10135f5c0 - middle::resolve::Resolver<'a>::resolve_implementation::closure.67222
  13:        0x10132e4a2 - middle::resolve::Resolver<'a>::with_type_parameter_rib::h9764680d6cf0a5cfEsW::v0.11.0.pre
  14:        0x1013079f5 - middle::resolve::Resolver<'a>.Visitor<(*>::visit_item::h8b1e7ccb2dd9f731wnR::v0.11.0.pre
  15:        0x10135304f - visit::Visitor::visit_mod::h2842851125407229744::v0.11.0.pre
  16:        0x10137408f - middle::resolve::resolve_crate::h647ea4130a3795acMnZ::v0.11.0.pre
  17:        0x10192a229 - driver::driver::phase_3_run_analysis_passes::ha0ab5f4b406c3845eGi::v0.11.0.pre
  18:        0x101924a2c - driver::driver::compile_input::haef4dbe816a3a3b5Ovi::v0.11.0.pre
  19:        0x1019c9f22 - driver::run_compiler::h4ac009c6e4455c14y7k::v0.11.0.pre
  20:        0x1019c7f1d - driver::main_args::closure.93216
  21:        0x1019de4d7 - driver::monitor::closure.94392
  22:        0x1019d8c4b - task::TaskBuilder::try::closure.94159
  23:        0x1010dcadc - task::spawn_opts::closure.8022
  24:        0x103fa12a8 - rt::task::Task::run::closure.25589
  25:        0x103fd3c7c - rust_try
  26:        0x103fa1211 - rt::task::Task::run::h66725e9dd8f1993fq1w::v0.11.0.pre
  27:        0x1010dc93f - task::spawn_opts::closure.7994
  28:        0x103fa90c6 - rt::thread::thread_start::h14fb024f55211441TKx::v0.11.0.pre
  29:     0x7fff92309899 - _pthread_body
  30:     0x7fff9230972a - _pthread_struct_init

cc @Ryman, this is in the new resolve code for better error suggestions.

[Ryman]

@alexcrichton thanks, I expected some unhandled cases just didn't really know what would trigger them :)

I think additional tests are needed for resolve errors inside impls of:

• *T -- TyPtr
• &T -- TyRPtr?
• Box<T> -TyBox (/ TyUniq?)
  These should be trivial to add.

I'm less sure about how to manage resolve suggestions for these, but having examples of impls for them would be neat to test with:

• Vec<T> - TyVec? or is this covering &[...]? I would expect Vec<T> to be a TyPath?
• [T,.. N] - TyFixedLenthVec?
• (T, U) - TyTup?
• TyBareFn/Proc/Closure

I'm assuming that we can't ever have an impl for TyNil, TyBot, TyTypeOf, TyInfer, or if we have, it's extremely specialised code that can probably do without specialised resolve errors.

[alexcrichton]

It's probably fine to just bail out early on, adding methods to a non-enum or non-struct is pretty rare

[Ryman]

Agreed, I'll test & fix the first three tomorrow or Sunday, and probably leave a note about the rest for the future.

[alexcrichton]

Awesome, thanks!