Crash after migrating from 0.19.0 to 0.20.1

[yanshay]

When migrating my app from esp_hal 0.19.0 to 0.20.1 (esp-hal and all related libraries) I witness the crash below.
The migration involved really only minor changes to the code.

The app still mostly works fine (wifi, pn532, display on screen, etc.), then when I press the touch screen the crash occurs.

When looking up the exception I found it means Access To Coprocessor 0 When Disabled, isn't Coprocessor 0 the default? So surely not disabled?

Some more information I managed to dig is:
When I disable a certain line that calls into slint (an embedded ui framework) this doesn't take place.
The slint line referred to in the log (api.rs:42) is:
Self::new(physical_pos.x as f32 / scale_factor, physical_pos.y as f32 / scale_factor)
which looks really harmless, and is also code that's there for a couple of years now, it's a library that worked for me with mostly the exact same code also on older versions of slint and older versions of esp-hal.

And again, the code works perfectly well with 0.19.0, the only thing I did is upgrade to 0.20.1 (and all other related libraries), this is why I'm asking here, since the exception seems esp32 related and the change to 0.20.1 caused it to occur.

Any ideas where to look next?

Exception occured 'Cp0Disabled'
Context
PC=0x42004bd9       PS=0x00060930
0x42004bd9 - i_slint_core::api::LogicalPosition::from_physical
    at /Users/my_user/.cargo/registry/src/index.crates.io-6f17d22bba15001f/i-slint-core-1.7.2/api.rs:42
A0=0x82049e57       A1=0x3fcdbe40       A2=0x3fc97a00       A3=0x3fc98308       A4=0x00000000
0x3fc97a00 - nfc_filament::__embassy_main::POOL
    at ??:??
0x3fc98308 - nfc_filament::__embassy_main::POOL
    at ??:??
A5=0x00000048       A6=0x0000005d       A7=0x3f800000       A8=0x00000001       A9=0x3fcdc590
A10=0x3f800000      A11=0x00000000      A12=0x00000000      A13=0x3fc90058      A14=0x3b9aca00
0x3fc90058 - nfc_filament::init_heap::HEAP
    at ??:??
A15=0x00000000
SAR=00000012
EXCCAUSE=0x00000020 EXCVADDR=0x00000000
LBEG=0x40056f5c     LEND=0x40056f72     LCOUNT=0x00000000
THREADPTR=0x00000000
SCOMPARE1=0x00000100
BR=0x00000001
ACCLO=0x00000000    ACCHI=0x00000000
M0=0x00000000       M1=0x00000000       M2=0x00000000       M3=0x00000000
F64R_LO=0x00000000  F64R_HI=0x00000000  F64S=0x00000000
FCR=0x00000000      FSR=0x00000003
F0=0x00000004       F1=0x3fcdc098       F2=0x3fc900a8       F3=0x3fc8fb60       F4=0x00010000
0x3fc900a8 - nfc_filament::init_heap::HEAP
    at ??:??
0x3fc8fb60 - nfc_filament::init_heap::HEAP
    at ??:??
F5=0x008301f8       F6=0xffffff00       F7=0x3fffffff       F8=0x00f80000       F9=0x42000183
F10=0x00000000      F11=0x00000404      F12=0x005d0048      F13=0xfff00100      F14=0xff04ffff
F15=0x00ff0000
0x420160a0
esp_hal_embassy::executor::thread::Executor::run
    at /Users/my_user/.cargo/registry/src/index.crates.io-6f17d22bba15001f/esp-hal-embassy-0.3.0/src/executor/thread.rs:126
0x4200e9ca
main
    at ??:??
0x4205f9db
Reset
    at ??:??
0x40378f57
ESP32Reset
    at ??:??

[bjoernQ]

Please try cargo update to make sure it picks up xtensa-lx-rt v0.17.1

If it still doesn't work, I guess that floating-point operation is done from within an interrupt handler? In that case see #2050 (comment)

[yanshay]

What are interrupt executors?
I'm using irq.wait_for_high() and irq.wait_for_low(), once these trigger I read the data arriving from I2C and continue.
it looks like this:

 let _ = with_timeout(Duration::from_millis(200), self.irq.wait_for_high()).await;

and at another place a simpler:

let _ = self.irq.wait_for_low().await;

So practically, I just wait for a pin, I guess underneath it maps to an interrupt.

[Dominaezzz]

Apologies, I didn't read your initial post thoroughly. This is what I was talking about https://github.com/esp-rs/esp-hal/blob/v0.20.1/esp-hal-embassy/src/executor/interrupt.rs#L25, and you're probably not using it.

[yanshay]

No, I'm not using it. At least not that I'm aware of.

[bugadani]

Interesting, can you please share your project, or a stripped-down version of it that still crashes?

[yanshay]

Interesting, can you please share your project, or a stripped-down version of it that still crashes?

I can try to reproduce it in a new project.

Given the discussion above and the solution that worked, I think the key things there would be using embassy and initializing wifi, following that await a pin using the code I showed above, and once that pin change make some floating point calculation.