Weird issue with two u8 ranged loops about avr-hal

[CoolSlimbo]

That's what it could be.

I guess I might fill my night with some figuring out a) what tf AVR assembly looks like, and b) figuring out how it results.

Fell free to close the issue, but I'll post updates on what I find.

[CoolSlimbo]

Rocking avr-gcc 14.1.0.

So could be a difference between platforms building on, or our avr-gcc versions, as it seems.

[CoolSlimbo]

Hey, just produced a minimum version of what you provided, can you attempt to run this (tested on an Arduino UNO)

#![no_std]
#![no_main]

use arduino_hal::prelude::*;
use panic_halt as _;

#[arduino_hal::entry]
fn main() -> ! {
    let dp = arduino_hal::Peripherals::take().unwrap();
    let pins = arduino_hal::pins!(dp);
    let mut serial = arduino_hal::default_serial!(dp, pins, 57600);

    loop {
        for x in (0..=255).rev() {
            ufmt::uwriteln!(&mut serial, "L1 {}", x).unwrap_infallible();
            arduino_hal::delay_ms(10);
        }
        for x in 0..=255 {
            ufmt::uwriteln!(&mut serial, "L2 {}", x).unwrap_infallible();
            arduino_hal::delay_ms(10);
        }
    }
}

(Created from a new project generated with cargo generate --git https://github.com/Rahix/avr-hal-template.git)

[burumdev]

Hi @CoolSlimbo .

Good idea. I tried commenting out everything as you suggested and it started working fine. But as soon as I activate the line that sets red LED's duty: led_rgb.red_output.set_duty(x); it starts to get stuck in the second loop again.

Then I comment the red LED line and activate green and blue LED lines, then everything is fine. Loops are alternating.

One difference between red LED and the others is that red uses Timer2Pwm and others use Timer0Pwm.

I also tried to use the other pin d3 or PD3 that Timer2Pwm uses and result is the same. Loop gets stuck with Timer2Pwm and set_duty().

One thing I should also add is the red LED set_duty line in the first loop doesn't cause any problems. If I comment out the second and the first remains, then the loops still alternate as expected.

For anyone interested I post the complete code that triggers the issue down below:

#![no_std]
#![no_main]

use arduino_hal::prelude::_unwrap_infallible_UnwrapInfallible;
use arduino_hal::hal::port::{ PB3, PD5, PD6 };
use arduino_hal::simple_pwm::{ Timer0Pwm, Timer2Pwm, Prescaler, IntoPwmPin };
use arduino_hal::port::{ Pin, mode };

use panic_halt as _;

struct LedRGB {
    red_output: Pin<mode::PwmOutput<Timer2Pwm>, PB3>,
    green_output: Pin<mode::PwmOutput<Timer0Pwm>, PD6>,
    blue_output: Pin<mode::PwmOutput<Timer0Pwm>, PD5>,
}

#[arduino_hal::entry]
fn main() -> ! {
    let dp = arduino_hal::Peripherals::take().unwrap();
    let pins = arduino_hal::pins!(dp);
    let timer0 = Timer0Pwm::new(dp.TC0, Prescaler::Prescale64);
    let timer2 = Timer2Pwm::new(dp.TC2, Prescaler::Prescale64);

    let mut serial = arduino_hal::default_serial!(dp, pins, 57600);

    let mut led_rgb = LedRGB {
        red_output: pins.d11.into_output().into_pwm(&timer2),
        green_output: pins.d6.into_output().into_pwm(&timer0),
        blue_output: pins.d5.into_output().into_pwm(&timer0),
    };

    led_rgb.red_output.enable();
    led_rgb.green_output.enable();
    led_rgb.blue_output.enable();

    let delay_ms = 10;

    loop {
        for x in (0..=255).rev() {
            led_rgb.red_output.set_duty(x);
            led_rgb.green_output.set_duty(255 - x);
            let blue: u8 = if x >= 128 { 128 - (255 - x) } else { 127 - x };
            led_rgb.blue_output.set_duty(blue);

            ufmt::uwriteln!(&mut serial, "LOOP 1: RED: {}, GREEN: {}, BLUE: {}", x, 255 - x, blue).unwrap_infallible();

            arduino_hal::delay_ms(delay_ms);
        }

        // This loop gets stuck and loops forever, but the first loop and this one should alternate
        for y in 0..=255 {
            // Commenting out below line will make the code work as expected.
            led_rgb.red_output.set_duty(y);
            // LEDs connected to Timer0Pwm work fine. Loops will alternate when I use them
            led_rgb.green_output.set_duty(255 - y);
            let blue: u8 = if y >= 128 { 128 - (255 - y) } else { 127 - y };
            led_rgb.blue_output.set_duty(blue);

            ufmt::uwriteln!(&mut serial, "LOOP 2: RED: {}, GREEN: {}, BLUE: {}", y, 255 - y, blue).unwrap_infallible();

            arduino_hal::delay_ms(delay_ms);
        }
    }
}

[CoolSlimbo]

Have you tried with a different timer to Timer2?

[burumdev]

Trying with Timer1Pwm it has the same bug. Tried changing prescaler values to 8, 256, 1024 etc. no effect there either.

[burumdev]

Though I tried one more thing and got interesting result:

        for y in 0..=255 {
            //led_rgb.red_output.set_duty(y);
            // LEDs connected to Timer0Pwm work fine. Loops will alternate when I use them
            led_rgb.green_output.set_duty(y);
            let blue: u8 = if y >= 128 { 128 - (255 - y) } else { 127 - y };
            led_rgb.blue_output.set_duty(blue);

            ufmt::uwriteln!(&mut serial, "LOOP 2: RED: {}, GREEN: {}, BLUE: {}", y, 255 - y, blue).unwrap_infallible();

            arduino_hal::delay_ms(delay_ms);
        }

led_rgb.green_output.set_duty(y); I changed this line to set the duty cycle directly to y counter value on green_output which uses Timer0Pwm and the bug appeared again.

So at this point setting the duty cycle to 255 directly after an ascending ramp of values triggers the bug, regardless of the PWM source for that pin. Very strange.

[CoolSlimbo]

That is. I'll have to dig around to find my RGB LED, and I'm about to help with another issue. I'll take your last full posted example and see what I can make of it on my Arduino and LED.

Also, for future code snippets, please add rs after the backticks to highlight it in Rust syntax. (I.e. ```rs)

[burumdev]

Then what I did shocked the civilized world:

            led_rgb.green_output.set_duty(255 - y);
            let blue: u8 = if y >= 128 { 128 - (255 - y) } else { 127 - y };
            led_rgb.blue_output.set_duty(blue);
            led_rgb.red_output.set_duty(y);

I restored the code to it's original bugging version, and simply moved led_rgb.red_output.set_duty(y); to the end of the lines that set the RGB led duty cycles and VOILA! It started to work as expected! Hope this helps find a solution. It looks like a compiler bug creating bad machine code to me.

Thank you for looking into this. Cheers!

[CoolSlimbo]

So, I attempted to run your code provided here, however, mine seemed to have run fine, without any issue, looping back and fourth between the two loops successfully.

So, to determine the issue that may be present, can I ask what toolchain version you are running on, and what commit your arduino_hal is pinned to, in your Cargo.toml?

[burumdev]

Sure. Here it is:

[dependencies.arduino-hal]
git = "https://github.com/rahix/avr-hal"
rev = "3e362624547462928a219c40f9ea8e3a64f21e5f"
features = ["arduino-uno"]

And rust-toolchain.toml content is:

[toolchain]
channel = "nightly-2024-03-22"
components = ["rust-src"]
profile = "minimal"

[CoolSlimbo]

Okay... very confused now.

The code example you said wasn't working, works fine for me on my UNO, and we have matching HAL's and toolchains...

This could've just been a single time fluke of some form with your compiler being special. Since I can't reproduce on a fresh install, and you've resolved it yourself, I don't see a reason as to why we have to keep this open.

If anyone has this issue in the future, and comes to this issue, feel free to open it up again.

(Free to close it now buru)

[burumdev]

Strange indeed. I also tried the same code with an original UNO board and result is the same. I'll close this issue now as it seems related to my GCC and GCC-AVR setup.

My GCC version is 10.2.1 and GCC-AVR is 5.4.0+Atmel3.6.2 on a Debian 11 system. So if anyone else bumps into this bug these might help. Finally can you post the versions of your GCC and GCC-AVR packages? Thanks a lot.

[burumdev]

I finally updated my Debian system from version 11 to 12. Which comes with:

GCC version 12.2
avr-gcc version 5.4.0+Atmel3.6.2 (Same as debian 11 version)

This update totally resolved the issue. So the bug was probably related to GCC version. If anybody comes across this behavior of loops not conforming, I recommend updating the development environment.

Thanks.