errata_xxx functions in system_nrf52840.c are optimized away

[JonasJ]

Hi,

During a debug session I noticed that the following functions in system_nrf52840.c are optimized away by the compiler since they always return true.

The functions errata_36(), errata_66(), errata_136() all show the same pattern as below.

static bool errata_36(void)
{
    if (*(uint32_t *)0x10000130ul == 0x8ul){
        if (*(uint32_t *)0x10000134ul == 0x0ul){
            return true;
        }
        if (*(uint32_t *)0x10000134ul == 0x1ul){
            return true;
        }
        if (*(uint32_t *)0x10000134ul == 0x2ul){
            return true;
        }
    }
    
    return true;
}

Is this done on purpose or is this a bug in the errata code?

I'm using SDK 15.2.0.

Thanks.

[Håkon Alseth]

Hi,

 

These should be casted to 

volatile uint32_t *

 to ensure that this is not optimized away.

I will report this internally. Thanks for reporting this back to us!

*edit* Which compiler do you use? It can be useful for us to know and use for testing in the future.

Kind regards,

Håkon

[JonasJ]

Will that actually help anything? It will probably cause the compiler to actually read those memory locations but the function will always return true regardless of the if statements. My gut feeling is that the last return true should be a return false.

Thanks.

[Håkon Alseth]

Hi,

 

The side effect of this one returning true is that three events is cleared on boot up, but you are 100 % correct, as it should return false if none of them are set. What I thought you commented on is the casts to registers that are not volatile, sorry for misunderstanding. I'll add this to the internal report.

Kind regards,

Håkon