Get Miri working on ARM again

[joshlf]

It looks like Miri is currently broken on ARM. @saethlin has provided basic mentoring instructions: #2791 (comment).

Based on this comment, it appears that the bug was introduced between nightly-2022-11-13 and nightly-2022-11-14.

Original contents

In this commit of zerocopy, cargo miri test fails with the following error (edited for brevity - see the full log here):

$ cargo +nightly-2023-02-14 miri test --package zerocopy --target arm-unknown-linux-gnueabi --all-features -- --skip ui
...
running 50 tests
error: unsupported operation: can't call foreign function `llvm.arm.hint` on OS `linux`
  --> /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/../../stdarch/crates/core_arch/src/arm_shared/hints.rs:69:5
   |
69 |     hint(HINT_YIELD);
   |     ^^^^^^^^^^^^^^^^ can't call foreign function `llvm.arm.hint` on OS `linux`
   |
   = help: this is likely not a bug in the program; it indicates that the program performed an operation that the interpreter does not support
   = note: BACKTRACE:
   = note: inside `std::arch::arm::__yield` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/../../stdarch/crates/core_arch/src/arm_shared/hints.rs:69:5: 69:21
   = note: inside `std::hint::spin_loop` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/hint.rs:204:22: 204:49
   = note: inside `std::sync::mpmc::utils::Backoff::spin_light` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/std/src/sync/mpmc/utils.rs:114:13: 114:37
   = note: inside `std::sync::mpmc::list::Channel::<tests::test::event::CompletedTest>::start_send` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/std/src/sync/mpmc/list.rs:250:21: 250:41
   = note: inside `std::sync::mpmc::list::Channel::<tests::test::event::CompletedTest>::send` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/std/src/sync/mpmc/list.rs:402:17: 402:39
   = note: inside `std::sync::mpmc::Sender::<tests::test::event::CompletedTest>::send` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/std/src/sync/mpmc/mod.rs:128:41: 128:61
   = note: inside `std::sync::mpsc::Sender::<tests::test::event::CompletedTest>::send` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/std/src/sync/mpsc/mod.rs:614:9: 614:27
   = note: inside `tests::test::run_test_in_process` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/test/src/lib.rs:693:5: 693:29
   = note: inside closure at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/test/src/lib.rs:573:39: 5[81](https://github.com/google/zerocopy/actions/runs/4197219373/jobs/7279259385#step:11:82):14
   = note: inside closure at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/test/src/lib.rs:600:37: 600:79
   = note: inside `std::sys_common::backtrace::__rust_begin_short_backtrace::<[closure@tests::test::run_test::run_test_inner::{closure#1}], ()>` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/std/src/sys_common/backtrace.rs:121:18: 121:21
   = note: inside closure at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/std/src/thread/mod.rs:558:17: 558:78
   = note: inside `<std::panic::AssertUnwindSafe<[closure@std::thread::Builder::spawn_unchecked_<'_, '_, [closure@tests::test::run_test::run_test_inner::{closure#1}], ()>::{closure#1}::{closure#0}]> as std::ops::FnOnce<()>>::call_once` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/panic/unwind_safe.rs:271:9: 271:19
   = note: inside `std::panicking::r#try::do_call::<std::panic::AssertUnwindSafe<[closure@std::thread::Builder::spawn_unchecked_<'_, '_, [closure@tests::test::run_test::run_test_inner::{closure#1}], ()>::{closure#1}::{closure#0}]>, ()>` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/std/src/panicking.rs:4[83](https://github.com/google/zerocopy/actions/runs/4197219373/jobs/7279259385#step:11:84):40: 483:43
   = note: inside `std::panicking::r#try::<(), std::panic::AssertUnwindSafe<[closure@std::thread::Builder::spawn_unchecked_<'_, '_, [closure@tests::test::run_test::run_test_inner::{closure#1}], ()>::{closure#1}::{closure#0}]>>` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x[86](https://github.com/google/zerocopy/actions/runs/4197219373/jobs/7279259385#step:11:87)_64-unknown-linux-gnu/lib/rustlib/src/rust/library/std/src/panicking.rs:447:19: 447:81
   = note: inside `std::panic::catch_unwind::<std::panic::AssertUnwindSafe<[closure@std::thread::Builder::spawn_unchecked_<'_, '_, [closure@tests::test::run_test::run_test_inner::{closure#1}], ()>::{closure#1}::{closure#0}]>, ()>` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/std/src/panic.rs:140:14: 140:33
   = note: inside closure at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/std/src/thread/mod.rs:557:30: 559:16
   = note: inside `<[closure@std::thread::Builder::spawn_unchecked_<'_, '_, [closure@tests::test::run_test::run_test_inner::{closure#1}], ()>::{closure#1}] as std::ops::FnOnce<()>>::call_once - shim(vtable)` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/ops/function.rs:250:5: 250:71
   = note: inside `<std::boxed::Box<dyn std::ops::FnOnce()> as std::ops::FnOnce<()>>::call_once` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/alloc/src/boxed.rs:19[88](https://github.com/google/zerocopy/actions/runs/4197219373/jobs/7279259385#step:11:89):9: 1[98](https://github.com/google/zerocopy/actions/runs/4197219373/jobs/7279259385#step:11:99)8:52
   = note: inside `<std::boxed::Box<std::boxed::Box<dyn std::ops::FnOnce()>> as std::ops::FnOnce<()>>::call_once` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/alloc/src/boxed.rs:1988:9: 1988:52
   = note: inside `std::sys::unix::thread::Thread::new::thread_start` at /home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/std/src/sys/unix/thread.rs:[108](https://github.com/google/zerocopy/actions/runs/4197219373/jobs/7279259385#step:11:109):17: 108:64

error: aborting due to previous error

error: test failed, to rerun pass `-p zerocopy --lib`

Caused by:
  process didn't exit successfully: `/home/runner/.rustup/toolchains/nightly-2023-02-14-x86_64-unknown-linux-gnu/bin/cargo-miri runner /home/runner/work/zerocopy/zerocopy/target/miri/arm-unknown-linux-gnueabi/debug/deps/zerocopy-51d03519d8186c51 --skip ui` (exit status: 1)
test alloc_support::tests::test_extend_vec_zeroed ... 
Error: Process completed with exit code 1.

[saethlin]

Ouch, sounds like we don't fully support the new standard library channels anymore. That's bad.

[oli-obk]

Why is this going through an arm path on x86_64-unknown-linux-gnu?

[bjorn3]

Because OP used --target to run on the arm target.

[oli-obk]

oh right, I just looked at the path in the error.

Yea, we don't test platforms other than x86_64/i686 x linux/windows/mac. Though with the usage that arm is getting in practice, we may want to start testing that, too

[joshlf]

Any guesses as to when this might have regressed? We pin to a particular nightly toolchain in our CI, so if there's an older non-buggy Miri, we could pin to that to unblock ourselves.

[oli-obk]

I don't know if we ever supported channel on arm, is this the first time you're running miri on those tests?

[joshlf]

Nope, been running them for a few months. The commit in question just updates what pinned toolchain versions we use in CI (before this commit, it was 1.61.0, 1.64.0, and nightly-2022-10-17).

[oli-obk]

yea just saw that. I guess without us testing against arm, such regressions may randomly happen to you again in the future even if we fix this one. I don't know whether there will be more problems if we just turn on tests for an arm target, but we can try.

[joshlf]

If it's not too much work to test on arm, that'd be great for us! But of course asking for free labor is always easy 😛

[saethlin]

It's not a huge amount of work. This might even be something for a new contributor to do.

• Clone the repo
• Run ./miri toolchain
• Run MIRI_TEST_TARGET=arm-unknown-linux-gnueabi ./miri test
• Figure out how to get that to pass, sounds like the situation in this issue in particular just needs a no-op shim but I also tried this and I see a type mismatch which is confusing... maybe someone who knows the target is less confused by it.
• Add this target to the bottom of ci.sh

[RalfJung]

If that all works, the target should also be added in https://github.com/rust-lang/miri#supported-targets.

[joshlf]

I posted this issue to This Week in Rust's Call for Participation. Hopefully we'll get someone to take a stab at it!

[joshlf]

I wasn't able to get git bisect on this repository, but I was able to use zerocopy's CI to bisect on nightly releases, and found that the bug was introduced between nightly-2022-11-13 and nightly-2022-11-14 (see here for more details). Hopefully that helps narrow things down!

[RalfJung]

Do you have git commit SHAs corresponding to these nightlies?

[LevitatingLion]

I saw your Call for Participation on This Week in Rust, would love to take a stab at this! I've never looked at Miri's code before, but this seems like a good opportunity to start.

[joshlf]

@RalfJung I'm not sure where that information lives; I was hoping someone on this thread might know.

@LevitatingLion Thanks!!

[RalfJung]

rust --version should show the commit ID. Usually this is somewhere in the CI logs, but I guess it depend on the setup.

[joshlf]

Ah gotcha. Here's the output:

$ rustc +nightly-2022-11-13 --version
rustc 1.67.0-nightly (6284998a2 2022-11-12)
$ rustc +nightly-2022-11-14 --version
rustc 1.67.0-nightly (e631891f7 2022-11-13)