Compilation ICE on nphysics crate

[trolleyman]

When trying to compile nphysics on a nightly created on or after 2016-04-14, the compiler errors. Before this date it is fine.

Example code:

git clone https://github.com/sebcrozet/nphysics
rustup update nightly-2016-04-13 && rustup run nightly-2016-04-13 cargo build # Works fine
rustup update nightly-2016-04-14 && rustup run nightly-2016-04-13 cargo build # Errors

Presumably was this commit that broke things, as it's a commit on that day that touches trans.

On Windows 64 bit GNU this is the output:

   Compiling nphysics3d v0.2.2 (https://github.com/sebcrozet/nphysics.git#b379109f)
error: internal compiler error: ../src/librustc_trans\debuginfo\type_names.rs:155: debuginfo: Trying to create type name for unexpected type: <na::Pnt3<f32> as nc::ncollide_math::Point>::Vect
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 'Box<Any>', ../src/libsyntax\errors/mod.rs:575

[alexcrichton]

cc @eddyb

[eddyb]

You sure it's that innocent commit and not, you know, my changes to debuginfo, for MIR?
Specifically, this normalization is gone, that could cause problems.

[eddyb]

@trolleyman I can't reproduce on master (just rebased one of my branches containing unrelated changes, so I had a compiler to check with).

EDIT: Still no reproduction, even after git checkout b379109f.

[trolleyman]

Ok so I was misinterpreting which crate was being compiled. nphysics compiles fine, but my crate, which uses nphysics, doesn't. I'm trying to narrow down the code now.

[trolleyman]

Here's some of the rust log that I got when compiling on debug builds:

INFO:rustc_trans::common: Cache hit: Binder(<nc::shape::Plane<na::Vec3<f32>> as nc::ray::RayCast<na::Pnt3<f32>, na::Iso3<f32>>>)
INFO:rustc::middle::region: CodeExtent(118736) = CallSiteScope { fn_id: 5171551, body_id: 5171571 } [parent=0]
INFO:rustc::middle::region: CodeExtent(118737) = ParameterScope { fn_id: 5171551, body_id: 5171571 } [parent=118736]
INFO:rustc::middle::region: CodeExtent(118738) = Misc(5171554) [parent=0]
INFO:rustc::middle::region: CodeExtent(118739) = Misc(5171557) [parent=0]
INFO:rustc::middle::region: CodeExtent(118740) = Misc(5171561) [parent=0]
INFO:rustc::middle::region: CodeExtent(118741) = Misc(5171566) [parent=0]
INFO:rustc::middle::region: CodeExtent(118742) = DestructionScope(5171571) [parent=118737]
INFO:rustc::middle::region: CodeExtent(118743) = Misc(5171571) [parent=118742]
INFO:rustc::middle::region: CodeExtent(118744) = Misc(5171572) [parent=118743]
INFO:rustc::middle::region: CodeExtent(118745) = Misc(5171573) [parent=118744]
INFO:rustc::middle::region: CodeExtent(118746) = Misc(5171574) [parent=118745]
INFO:rustc::middle::region: CodeExtent(118747) = Misc(5171575) [parent=118745]
INFO:rustc::middle::region: CodeExtent(118748) = Misc(5171576) [parent=118745]
INFO:rustc::middle::region: CodeExtent(118749) = Misc(5171577) [parent=118745]
INFO:rustc::middle::region: CodeExtent(118750) = Misc(5171578) [parent=118744]
INFO:rustc::middle::region: CodeExtent(118751) = CallSiteScope { fn_id: 5171578, body_id: 5171582 } [parent=118750]
INFO:rustc::middle::region: CodeExtent(118752) = ParameterScope { fn_id: 5171578, body_id: 5171582 } [parent=118751]
INFO:rustc::middle::region: CodeExtent(118753) = Misc(5171580) [parent=0]
INFO:rustc::middle::region: CodeExtent(118754) = DestructionScope(5171582) [parent=118752]
INFO:rustc::middle::region: CodeExtent(118755) = Misc(5171582) [parent=118754]
INFO:rustc::middle::region: CodeExtent(118756) = Misc(5171583) [parent=118755]
INFO:rustc::middle::region: CodeExtent(118757) = Misc(5171584) [parent=118756]
INFO:rustc::middle::region: CodeExtent(118758) = DestructionScope(5171551) [parent=0]
INFO:rustc::middle::region: CodeExtent(118759) = CallSiteScope { fn_id: 11887, body_id: 11907 } [parent=1]
INFO:rustc::middle::region: CodeExtent(118759) = CallSiteScope { fn_id: 11887, body_id: 11907 } [parent=1] BOGUS!
INFO:rustc::middle::region: CodeExtent(118759) = CallSiteScope { fn_id: 11887, body_id: 11907 } [parent=1] BOGUS!
INFO:rustc::middle::region: CodeExtent(118759) = CallSiteScope { fn_id: 11887, body_id: 11907 } [parent=1] BOGUS!
INFO:rustc::middle::region: CodeExtent(118759) = CallSiteScope { fn_id: 11887, body_id: 11907 } [parent=1] BOGUS!
INFO:rustc::middle::region: CodeExtent(118759) = CallSiteScope { fn_id: 11887, body_id: 11907 } [parent=1] BOGUS!
INFO:rustc::middle::region: CodeExtent(11868) = Misc(11909) [parent=1] BOGUS!
INFO:rustc::middle::region: CodeExtent(11868) = Misc(11909) [parent=1] BOGUS!
INFO:rustc::middle::region: CodeExtent(11868) = Misc(11909) [parent=1] BOGUS!
INFO:rustc::middle::region: CodeExtent(11868) = Misc(11909) [parent=1] BOGUS!
INFO:rustc::middle::region: CodeExtent(11868) = Misc(11909) [parent=1] BOGUS!
INFO:rustc::middle::region: CodeExtent(11868) = Misc(11909) [parent=1] BOGUS!
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as std::marker::Sized>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as nc::ncollide_math::Scalar>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as std::marker::Sized>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as nc::ncollide_math::Scalar>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as std::marker::Sized>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as nc::ncollide_math::Scalar>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as std::marker::Sized>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as nc::ncollide_math::Scalar>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as std::marker::Sized>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as nc::ncollide_math::Scalar>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as std::marker::Sized>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as nc::ncollide_math::Scalar>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as std::marker::Sized>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as nc::ncollide_math::Scalar>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as std::marker::Sized>))`
INFO:rustc::traits::fulfill: check_duplicate_trait hit: `Binder(TraitPredicate(<f32 as nc::ncollide_math::Scalar>))`
error: internal compiler error: ../src/librustc_trans\debuginfo\type_names.rs:155: debuginfo: Trying to create type name for unexpected type: <na::Pnt3<f32> as nc::ncollide_math::Point>::Vect

[eddyb]

@trolleyman That's not really helpful, sadly, nightly doesn't use -C debug-assertions yet, so there's no debug! output.

[eddyb]

@trolleyman This bug is not in beta, correct?

[trolleyman]

@eddyb I'm afraid that I can't test that as I'm using #[feature]s.

[eddyb]

@trolleyman Can you check the test from #33099 instead? Particularly interested about the nightlies you had issues with.

[trolleyman]

@eddyb Sure, unfortunately that test case works fine for both pre and post-2016-04-14.

[eddyb]

@trolleyman Ah, beats me then.

[trolleyman]

@eddyb I'll try and find which part of my code hits the error.

[trolleyman]

@eddyb So I did some digging and I was wrong again, it turns out it is the nphysics crate. This is a minimal test case:

extern crate nphysics3d;

pub fn main() {
    nphysics3d::world::World::new().step(1.0);
}

[eddyb]

@trolleyman It depends on the nphysics3d crate, but it's not the crate itself, and that testcase is still large in terms of how much it needs to monomorphize.

[trolleyman]

@eddyb True, sorry but I am not knowledgeable at all in rustc or nphysics, so it's the best I can do :/

[nikomatsakis]

triage: P-high

[eddyb]

Still need a good regression test.

[nikomatsakis]

triage: P-low