I tried this code (also available at https://github.com/benhansen-io/esp-miscompile-test)
use esp_idf_sys as _; // If using the `libstart` feature of `esp-idf-sys`, always keep this module imported
#[derive(Copy, Clone, Debug, Default, Eq, PartialEq, Ord, PartialOrd, Hash)]
pub struct RGB {
r: u8,
g: u8,
b: u8,
}
impl RGB {
fn new(r: u8, g: u8, b: u8) -> RGB {
RGB { r, g, b }
}
}
fn next_color(curr: RGB, order: &[RGB]) -> RGB {
let curr_idx = order
.iter()
.position(|c| {
// println!("UNCOMMENT TO STOP PANIC");
*c == curr
})
.unwrap();
let next_idx = if curr_idx == order.len() - 1 {
0
} else {
curr_idx + 1
};
order[next_idx]
}
#[no_mangle]
extern "C" fn rust_main() -> i32 {
// Temporary. Will disappear once ESP-IDF 4.4 is released, but for now it is necessary to call this function once,
// or else some patches to the runtime implemented by esp-idf-sys might not link properly.
esp_idf_sys::link_patches();
let green: RGB = RGB::new(0, 255, 0);
let blue: RGB = RGB::new(0, 0, 255);
let yellow: RGB = RGB::new(255, 255, 0);
let color_order = [yellow, blue, green];
let mut curr = yellow;
println!("Starting");
loop {
println!("curr: {:?}", curr);
curr = next_color(curr, &color_order);
println!("next: {:?}", curr);
std::thread::sleep(std::time::Duration::from_millis(500));
}
}I expected to see this happen:
2023-02-15 20:44:15 Hello world from C!
2023-02-15 20:44:15 Starting
2023-02-15 20:44:15 curr: RGB { r: 255, g: 255, b: 0 }
2023-02-15 20:44:15 next: RGB { r: 0, g: 0, b: 255 }
2023-02-15 20:44:15 curr: RGB { r: 0, g: 0, b: 255 }
2023-02-15 20:44:16 next: RGB { r: 0, g: 255, b: 0 }
2023-02-15 20:44:16 curr: RGB { r: 0, g: 255, b: 0 }
2023-02-15 20:44:16 next: RGB { r: 255, g: 255, b: 0 }
2023-02-15 20:44:16 curr: RGB { r: 255, g: 255, b: 0 }
...
Instead, this happened:
2023-02-15 20:43:28 Hello world from C!
2023-02-15 20:43:28 Starting
2023-02-15 20:43:28 curr: RGB { r: 255, g: 255, b: 0 }
2023-02-15 20:43:28 next: RGB { r: 0, g: 0, b: 255 }
2023-02-15 20:43:28 curr: RGB { r: 0, g: 0, b: 255 }
2023-02-15 20:43:28 ESP-ROM:esp32s3-20210327
2023-02-15 20:43:29 curr: RGB { r: 0, g: 0, b: 255 }
2023-02-15 20:43:29 ESP-ROM:esp32s3-20210327
2023-02-15 20:43:30 curr: RGB { r: 0, g: 0, b: 255 }
2023-02-15 20:43:30 ESP-ROM:esp32s3-20210327
...
Without panic abort immediate I also see:
2023-02-15 20:49:27 thread '<unnamed>' panicked at 'called `Option::unwrap()` on a `None` value', src/lib.rs:23:10
Adding the commented out println! or removing the opt-level = "s" both cause the correct output to be displayed.
Meta
rustc --version --verbose:
rustc 1.65.0 (897e37553 2022-11-02)
binary: rustc
commit-hash: 897e37553bba8b42751c67658967889d11ecd120
commit-date: 2022-11-02
host: x86_64-unknown-linux-gnu
release: 1.65.0
LLVM version: 15.0.0
(tested on 1.67.0 too)
I tried this code (also available at https://github.com/benhansen-io/esp-miscompile-test)
I expected to see this happen:
Instead, this happened:
Without panic abort immediate I also see:
Adding the commented out println! or removing the
opt-level = "s"both cause the correct output to be displayed.Meta
rustc --version --verbose:(tested on 1.67.0 too)