Log printing right before reset causes broken text over RTT

[Mr_BLE]

Used SDK: 15.2.0

Hello,

I want to print some debug messages over RTT after an error occured. After that I want to reset the system.

If I put the string into the que with NRF_LOG_INFO("Debug Text"), then forcing the log module to print out all the strings in buffer with NRF_LOG_FINAL_FLUSH(), then calling NVIC_SystemReset() to reboot the system will cause broken text output in J-Link RTT viewer. It seems that the NRF_LOG_FINAL_FLUSH() function does not block the code until all the text was sended out.

Including a delay before triggering the reset will solve the problem. But with various length of the debug messages, calling a fixed delay is not a well made solution. Also the documentation sais, that calling NRF_LOG_FINAL_FLUSH() is made for exactly this problem.

Do you know whats going wrong?

Thank you, Regards

Mr_BLE

[catsunami]

It probably has to do with the limited buffer size of RTT. By default nrf_log uses SEGGER_RTT_MODE_NO_BLOCK_SKIP, which is probably why the text output is broken. You can try either increasing RTT buffer or setting SEGGER_RTT_MODE_DEFAULT to SEGGER_RTT_MODE_BLOCK_IF_FIFO_FULL.

[Mr_BLE]

Thank you for answering!

I tried both of your solutions, but none helped:

The first messages on top are with a delay inserted and the following messages are with your soltions:

- in sdk_config.h setting SEGGER_RTT_CONFIG_DEFAULT_MODE to "2" (BLOCK mode)

- in sdk_config.h double the buffers SEGGER_RTT_CONFIG_BUFFER_SIZE_DOWN and SEGGER_RTT_CONFIG_BUFFER_SIZE_UP

Do you have any others ideas?

Thank you.
Regards

[Jared]

Hi Andre,

Could you try:

• Disabling interrupts before calling NRF_LOG_FINAL_FLUSH():

    __disable_irq();
    NRF_LOG_FINAL_FLUSH();

• Replacing NRF_LOG_FINAL_FLUSH with NRF_LOG_FLUSH()

Jared 

[Mr_BLE]

Hi Jared,

sadly, it does not help. The output is still scrappy...

Regards

[Krzysztof Chruscinski]

Hi,

unfortunately, there is no RTT API to check if data is read by the host (PC) so there are two options: 

- arbitrary delay. Something like 10ms should be enough as RTT reads the buffer every couple milliseconds.

- software breakpoint after flush (see NRF_BREAKPOINT_COND in app_util_platform.h) which will break allowing host to read the buffer. In that case device will not reset.