Debug BLE "Softdevice"

[dilbert168]

I recently got the nRF52-DK for using as BLE Relay evaluation.

I have downloaded following and installed to run the code

- Segger Embedded Studio (v5.1)

- nRF5_SDK_17.0.2_d674dde

1. I have searched the forum and others saying debugger cannot be used when "Stop" and "Continue"

as BLE Stack is using same CPU to run.  Is this still the limitation now ?

2. Also, the local variable cannot be added to watch, is this limitation in Segger ?

[awneil]

dilbert168 said:

debugger cannot be used when "Stop" and "Continue" as BLE Stack is using same CPU to run.  Is this still the limitation now ?

Yes - you still need the Monitor Mode Debug.

dilbert168 said:

the local variable cannot be added to watch

That's not true.

However - as for any modern compiler and any IDE - you have to beware of compiler optimisation. This is a standard 'C' programming thing; not specific to Nordic or Segger - or even embedded.

[dilbert168]

my mistake , I meant to refer to "static variable".

In below example : 

static ble_evt_t const *p_local_evt;

static void ble_evt_handler(ble_evt_t const * p_ble_evt, void * p_context)
{
      // For readability.
      p_local_evt = p_ble_evt;
}

The "p_local_evt" will not able to perform "Quick Watch".  But it works if I removed "static".

I know this has nothing to do with Nordic or firmware.

Just wonder this is known issue in Segger as well... As I am using TI (CCStudio) has same issue.

[Kenneth]

Typically if you don't want the compiler to optimize away the variable you can use volatile when declaring the variable;

https://stackoverflow.com/questions/246127/why-is-volatile-needed-in-c 

Kenneth

[awneil]

You need to show a complete example which illustrates the issue.

If p_local_evt  is static, and is never read, then it is certainly up for optimisation.

By removing static, you make it global; so the compiler cannot know if it is referenced elsewhere or not - so cannot optimise it away.

[awneil]

I don't think it's guaranteed that volatile will stop the compiler from removing an entirely useless variable?

See: www.avrfreaks.net/.../3004761