qemu_riscv32/qemu_virt_riscv32/smp sample.rust/bench.plain fail

[fabiobaltieri]

Describe the bug
Hi, bumped into a couple issue with bench.plain tests in CI on main: https://github.com/zephyrproject-rtos/zephyr/runs/37777088940

Then I tried to reproduce it locally but I hit one more issues and can't even build it at all

95    |                 zephyr_sys::k_str_out(self.buf.as_mut_ptr() as *mut i8, self.count);
      |                 --------------------- ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected `*mut u8`, found `*mut i8`

confirmed by other users after updating the compiler.

@d3zd3z could you look into these?

To Reproduce
west build -p -b qemu_cortex_m3 ../modules/lang/rust/samples/bench

Expected behavior
build and run

Impact
CI is sad. Rust users are also sad I guess.

Environment (please complete the following information):
Linux, Zephyr SDK, a718b84

$ rustup --version
rustup 1.26.0 (2024-08-30)
info: This is the version for the rustup toolchain manager, not the rustc compiler.
info: The currently active `rustc` version is `rustc 1.85.0 (4d91de4e4 2025-02-17)`

[fabiobaltieri]

cc @kartben @pdgendt

[kartben]

It builds fine for me with rustc 1.84.0 but I can confirm the deadlock/timeout on the smp target, while the non-SMP variant seem to run the test just fine.

[pdgendt]

I think it might be related to rust-lang/rust#132975 as mentioned in the 1.85.0 release notes:

Make core::ffi::c_char signedness more closely match that of the platform-default char

This changed c_char from an i8 to u8 or vice versa on many Tier 2 and 3 targets (mostly Arm and RISC-V embedded targets). The new definition may result in compilation failures but fixes compatibility issues with C.

The libc crate matches this change as of its 0.2.169 release.

[fabiobaltieri]

I think it might be related to rust-lang/rust#132975 as mentioned in the 1.85.0 release notes:

How's that going to work in CI? I'd imagine that right now we are pinned to whatever version was there when the image was generated but then if it gets regenerated with the same config it would fall apart.

[pdgendt]

How's that going to work in CI? I'd imagine that right now we are pinned to whatever version was there when the image was generated but then if it gets regenerated with the same config it would fall apart.

It depends on the rust version on the ubuntu image.

[fabiobaltieri]

@d3zd3z should this not run on the SMP targets? Sounds like that combination is not stable, probably a better idea to pick some explicit targets and that's it.