detecting memory issues at runtime

[paul_tanner]

I am using nrf52 with s132 and armgcc As I add functionality to the code I get to a stage where it fails. Linker runs OK but the result fails to execute.

The observable phenomenon is that I see nothing with Jlink, not even the first printf at the top of main.c

I can verify that it's a memory issue by changing the amount of ram reserved for static arrays.

My question is whether I can trap such things as stack overflow and get some information on what's happening?

I also need to analyse the build to identify any memory hogs. Any tools for that would be useful.

Thx.

[Hung Bui]

The linker should be able to detect static memory over flow. Do you have any example that reproduce the issue ? When debugging, have you checked if PC can actually enter the main loop ?

[paul_tanner]

My example is big and complex. I'll make a test using simple code and a large static array. To check if it enters the main loop I was using RTT printfs. As noted above, the first of these does not show on JLinkRTTClient so it probably crashes before that. I think this could be due to stack/ heap overflow.

[Hung Bui]

Using RTT might not be the best option to debug this issue as RTT also requires a static RAM block to operate. Also if the log is deffered and the code get stuck before you call NRF_LOG_PROCESS() you won't get anything on RTT viewer.

[paul_tanner]

Makes sense thx. I'm not sure what would be the best approach. Could I get some other indication and light a LED?

[Hung Bui]

You can try to add a break point and step into the code, you can start with SystemInit() inside system_nrf52.c if you can't get to the main loop. Toggling an LED could also be a good way if behavior when debugging is different from normal running.