ADC read locked forever with CONFIG_DEBUG_OPTIMISATION=y on STM32U5

[tpennors]

Describe the bug
ADC read locked forever with CONFIG_DEBUG_OPTIMISATION=y on STM32U5
I used gdb to analyse this. I saw that the code gets locked in :
drivers/adc/adc_context.h: k_sem_take(&ctx->sync, K_FOREVER);
I checked and adc_stm32_isr is never called.

When using CONFIG_NO_OPTIMISATION=y, everything is fine.

Temporary fix
I found a temporary fix but I would like to understand if it is a real bug from the driver. My solution is to add a very little waiting period between those two calls: LL_ADC_EnableIT_EOC(adc); and adc_context_start_read(&data->ctx, sequence);

Other
I didn't use overlay as it is described in sample/drivers/adc/boards/stm32l562e_dk.overlay, would that be a prolem ?

[nordicjm]

Probably a stack overflow, increase the stack sizes and try again

[tpennors]

Hello,

I tried to increase stack size and still doesn't work. Hardware interrupt is never received. I would have think of a timing between writing to registers which is not the same between optimizations. Did anyone see this kind of behaviour ?

[FRASTM]

When building the samples for nucleo_u575zi_q with CONFIG_DEBUG_OPTIMIZATIONS=y
The measurement is correct on ADC1 channel 1 (available on Arduino connector CN9 pin A5)
Of course is the overlay required to test (try to connect pin A5 to 3.3v line)

ADC reading:                                                                                                                   
- adc@42028000, channel 1: 16383 = 3299 mV   

[legoabram]

@tpennors We might be running into a similar problem in Zephyr v3.5.0-rc3. Did you ever find a solid answer to this problem?