Our last ICE before we get coverage working

[gilescope]

Thanks for all the work you've recently been doing to coverage - we seem to have one last hurdle before all our tests (closed source) pass with coverage enabled. (yes we use proc macros)

RUSTCFLAGS = -Zinstrument-coverage -Zprofile -Ccodegen-units=1 -Copt-level=0 -Clink-dead-code -Coverflow-checks=off
CARGO_INCREMENTAL=0
RUSTDOCFLAGS=-Cpanic=abort

cargo test --message-format=json -Ztimings=json,html,info -- -Zunstable-options --format json

test failed
thread 'rustc' panicked at 'called `Option::unwrap()` on a `None` value', compiler/rustc_codegen_llvm/src/context.rs:328:35
stack backtrace:
0: rust_begin_unwind
at /rustc/043eca7f0b34d12e61c44206beca740628647080/library/std/src/panicking.rs:483:5
1: core::panicking::panic_fmt
at /rustc/043eca7f0b34d12e61c44206beca740628647080/library/core/src/panicking.rs:85:14
2: core::panicking::panic
at /rustc/043eca7f0b34d12e61c44206beca740628647080/library/core/src/panicking.rs:50:5
3: rustc_codegen_llvm::coverageinfo::<impl rustc_codegen_ssa::traits::coverageinfo::CoverageInfoBuilderMethods for rustc_codegen_llvm::builder::Builder>::add_counter_region
4: rustc_codegen_ssa::mir::block::<impl rustc_codegen_ssa::mir::FunctionCx<Bx>>::codegen_block
5: rustc_codegen_ssa::mir::codegen_mir
6: rustc_codegen_ssa::base::codegen_instance
7: <rustc_middle::mir::mono::MonoItem as rustc_codegen_ssa::mono_item::MonoItemExt>::define
8: rustc_codegen_llvm::base::compile_codegen_unit::module_codegen
9: rustc_query_system::dep_graph::graph::DepGraph<K>::with_task
10: rustc_codegen_llvm::base::compile_codegen_unit
11: rustc_codegen_ssa::base::codegen_crate
12: <rustc_codegen_llvm::LlvmCodegenBackend as rustc_codegen_ssa::traits::backend::CodegenBackend>::codegen_crate
13: rustc_session::utils::<impl rustc_session::session::Session>::time
14: rustc_interface::queries::Queries::ongoing_codegen
15: rustc_interface::queries::<impl rustc_interface::interface::Compiler>::enter
16: rustc_span::with_source_map
17: scoped_tls::ScopedKey<T>::set
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
error: internal compiler error: unexpected panic
note: the compiler unexpectedly panicked. this is a bug.
note: rustc 1.49.0-nightly (043eca7f0 2020-10-17) running on x86_64-unknown-linux-gnu
note: compiler flags: -C codegen-units=1 -C embed-bitcode=no -C panic=abort --crate-type bin
query stack during panic:
end of query stack
Couldn't compile the test.

[gilescope]

@richkadel might be interested in this?

rust/compiler/rustc_codegen_llvm/src/context.rs

Line 328 in 07a63e6

self.coverage_cx.as_ref().unwrap()

I don't know how much help this report is... but I appreciate all the work you've been doing.

[jyn514]

@gilescope you said the original code is private - could you put together a minimal reproduction?

[gilescope]

That's a very good question. Not sure - it's not that minimal at the moment.. on the plus side this issue is affecting 1% of the tests, so we're really very close to everything working...

tcx.sess.opts.debugging_opts.instrument_coverage - it looks like the code thought coverage was not turned on but then later it tried to log some coverage.

[richkadel]

My best guess is you have a library compiled with coverage and are compiling a binary without the -Zinstrument-coverage flag. The coverage statements in the library are encountered during codegen, which assumes coverage is enabled.

Assuming this is the scenario, this should be an easy fix.

I'll put a PR together.

FYI @wesleywiser @tmandry

[richkadel]

@gilescope I think the PR I just uploaded should fix this issue.

@wesleywiser - I hope you don't mind, I set you as the reviewer. OK?

[wesleywiser]

@richkadel Yep, totally fine!

[richkadel]

@gilescope The proposed fix has been merged. Can you rebase and retest and verify this resolves the issue?

[gilescope]

I'd love to be in a position to tell you if it works or not but because of reasons I can't give you a definitive answer one way or another at the moment. The changes look good to me though - much appreciated and apologies I can't give you a definite answer yet.

[richkadel]

Ok, I understand. Should we close this issue and reopen a new issue if the problem returns?

[gilescope]

Yes I think best close this - thanks very much for the fix.