No stack trace in app_error_fault_handler

RE: No stack trace in app_error_fault_handler

fkoran — Fri, 07 Jun 2019 21:19:25 GMT

I have the same issue. Segger 4.16, the DEBUG flag is set. Optimization level is set to 'debug'

RE: No stack trace in app_error_fault_handler

RK — Sat, 25 May 2019 02:43:49 GMT

That assembly doesn't push the frame pointer (r11) and then set it to point to the new stack location. That probably breaks the stack trace.

RE: No stack trace in app_error_fault_handler

Eliot Stock — Fri, 24 May 2019 12:46:52 GMT

I'm also seeing this.

RE: No stack trace in app_error_fault_handler

haakonsh — Mon, 12 Nov 2018 14:56:23 GMT

What IDE are you using? Have you set the DEBUG flag?

RE: No stack trace in app_error_fault_handler

abhipray — Wed, 07 Nov 2018 21:55:53 GMT

I am seeing the same issue using SDK 15.2. When APP_ERROR_CHECK triggers a call to app_error_handler(), the call stack is retained. Once it finishes executing the assembly inside app_error_handler() and enters app_error_fault_handler(), it loses the call stack. I have tested to see if it was a problem with app_error_fault_handler() by using the WEAK implementation and also by calling it directly in place of APP_ERROR_CHECK, and it doesn't appear to be the problem. It looks like the corruption is happening in the app_error_handler() assembly.

RE: No stack trace in app_error_fault_handler

haakonsh — Tue, 26 Jun 2018 08:44:33 GMT

I'm guessing the compiler optimizes the loop as it knows the while(true) check will always pass. When you introduce a volatile variable you tell the compiler not to optimize this piece of code as you expect the variable to change during runtime. The optimization yields code that is not debug friendly.

RE: No stack trace in app_error_fault_handler

A Knecht — Fri, 08 Jun 2018 15:33:43 GMT

I just found that the call stack doesn't disappear when I replace while(true) with volatile bool loop = true; while(loop)

This is analogous to line 111 in app_error.c (SDK 14.0.0).

// The following variable helps Keil keep the call stack visible, ...
volatile bool loop = true;

Can any Nordic employee share some insights as to why this is required and how it works?

RE: No stack trace in app_error_fault_handler

haakonsh — Fri, 04 May 2018 08:24:43 GMT

What happens if you copy the weak implementation into your main.c (overrides from main context) ?

RE: No stack trace in app_error_fault_handler

A Knecht — Thu, 03 May 2018 07:55:12 GMT

No, interestingly not. It's only when I use my implementation.

RE: No stack trace in app_error_fault_handler

haakonsh — Wed, 02 May 2018 15:07:06 GMT

Do you see the same issue if you use the weak implementation of app_error_fault_handler() in app_error_weak.c?

RE: No stack trace in app_error_fault_handler

A Knecht — Wed, 02 May 2018 14:24:57 GMT

APP_ERROR_FAULT_HANDLER(0)

(I'm causing a deliberate crash by putting this statement in the first line of the main function.) What's unexpected is that there is no stack trace.

RE: No stack trace in app_error_fault_handler

haakonsh — Wed, 02 May 2018 13:47:03 GMT

What's on line 0x81 of main.c?

RE: No stack trace in app_error_fault_handler

A Knecht — Mon, 30 Apr 2018 07:39:16 GMT

id: 0x4001
pc the one passed into the fault handler: 0x0
info: 0x2000dc00
the contents as expected: (0x81, '..\main.c' and 0)
R15 (pc) is: 0x27972
and lr: 0x279b3
location/value of the only item on the stack trace (app_error_fault_handler) also reads 0x0

RE: No stack trace in app_error_fault_handler

haakonsh — Thu, 26 Apr 2018 13:50:02 GMT

What's the value of id, pc, and info?

RE: No stack trace in app_error_fault_handler

A Knecht — Thu, 26 Apr 2018 09:28:28 GMT

Tested it, this didn't change anything.

RE: No stack trace in app_error_fault_handler

haakonsh — Tue, 24 Apr 2018 12:01:21 GMT

Hey Andreas,

Try calling APP_ERROR_CHECK(arg), where arg is 1-17 dec, instead of APP_ERROR_HANDLER(0) .

Cheers,

Håkon.