ICE in Creader::resolve_crate() when resolving unknown crate from metadata

[abonander]

rustc 1.17.0-nightly (6f10e2f63 2017-03-14)

Backtrace and last log statement before panic:

INFO:rustc_metadata::creader: resolving crate `extern crate rayon as rayon`
error: internal compiler error: unexpected panic

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: run with `RUST_BACKTRACE=1` for a backtrace

thread 'rustc' panicked at 'called `Option::unwrap()` on a `None` value', /checkout/src/libcore/option.rs:323
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
stack backtrace:
   0: rustc_metadata::creader::CrateLoader::resolve_crate
   1: <rustc_metadata::creader::CrateLoader<'a> as rustc::middle::cstore::CrateLoader>::process_item
   2: rustc_resolve::build_reduced_graph::<impl rustc_resolve::Resolver<'a>>::build_reduced_graph_for_item
   3: <rustc_resolve::build_reduced_graph::BuildReducedGraphVisitor<'a, 'b> as syntax::visit::Visitor<'a>>::visit_item
   4: <rustc_resolve::build_reduced_graph::BuildReducedGraphVisitor<'a, 'b> as syntax::visit::Visitor<'a>>::visit_item
   5: rustc_resolve::macros::<impl syntax::ext::base::Resolver for rustc_resolve::Resolver<'a>>::visit_expansion
   6: syntax::ext::expand::MacroExpander::collect_invocations
   7: syntax::ext::expand::MacroExpander::expand
   8: syntax::ext::expand::MacroExpander::expand_crate
   9: rustc_driver::driver::phase_2_configure_and_expand::{{closure}}
  10: rustc_driver::driver::phase_2_configure_and_expand
  11: rustc_driver::driver::compile_input
  12: rustc_driver::run_compiler
  13: std::panicking::try::do_call
  14: __rust_maybe_catch_panic
             at /checkout/src/libpanic_unwind/lib.rs:98
  15: <F as alloc::boxed::FnBox<A>>::call_box
  16: std::sys::imp::thread::Thread::new::thread_start
             at /checkout/src/liballoc/boxed.rs:650
             at /checkout/src/libstd/sys_common/thread.rs:21
             at /checkout/src/libstd/sys/unix/thread.rs:84
  17: start_thread
  18: clone

error: Could not compile `img-dup`.

To learn more, run the command again with --verbose.

I have the code that caused the ICE (including Cargo.lock) branched off here: https://github.com/abonander/img-dup/tree/ice_mar_2017

@jseyfried I'd like to work on this one, first step will be to attempt building with a debug-mode rustc to get the exact line number.

[jseyfried]

@abonander Great! Is this a regression?

[abonander]

Not sure, let me check.

[abonander]

@jseyfried Looks like a beta-to-nightly. On stable and beta I get a regular old "cannot find crate rayon" because it wasn't in my Cargo deps.

[jseyfried]

Thanks! Self-assigning so I don't forget about this.
My first step would be to figure out what PR caused this -- I'll do that this weekend unless you figure it out first :)

[abonander]

Seems it's a little more complicated than I first expected. I use cargo check frequently to make sure my project compiles without wasting time in trans, and it seems that that's significant, because cargo build works fine on nightly (correctly reports the unknown crate error). cargo check also ICEs on rustc 1.17.0-beta.1 (408d49e60 2017-03-14), but on stable it instead prints an error about invalid arguments to rustc.

So this may be due to argument parsing changes instead of changes in Creader, or an unfortunate interaction of such changes, but it also appears to be a stable-to-beta regression since stable doesn't ICE. I initially thought it was a beta-to-nightly regression because beta initially printed the same error as stable, but my beta was outdated and it ICEs after updating.

[abonander]

I guess it's back to metadata because the arguments don't appear malformed, cargo check is just having rustc emit rmeta's instead of rlib's: https://gist.github.com/abonander/a1b7df5cde0be49e1fc4cbf604771023

So... stable-beta regression of metadata loading, I guess?

[jseyfried]

cc @nrc

[TimNN]

Maybe related: #40469

[abonander]

Found the problematic unwrap() by replacing it with expect(), it was my first suspect too: https://github.com/rust-lang/rust/blob/master/src/librustc_metadata/creader.rs#L238

My first instinct is to just add an .or(source.rmeta.as_ref()) there but I don't think it's that simple; it shouldn't be getting this far with a nonexistent crate, should it?

[abonander]

Wow, I guess it was that simple. I don't get the ICE anymore. PR incoming.

[abonander]

Minimal reproduction case: https://github.com/abonander/rust-issue-40535