GCC compiles fuctions without return instruction

Ferdinand Fahlbusch Guest 2009-04-01 00:16 #1214492

Rate this post

•	▲ useful ▼ not useful

Hello.

I´m using an ARM from OLIMEX, the STM32-P103 with a Cortex-M3. I use 
eclipse to write my code in C, arm-none-eabi-gcc to compile, 
arm-none-eabi-gdb to debug and OpenOCD to interface the microcontroller 
trough an ARM-USB-OCD(based on FT2232L), all this in Windows.

My program compiles and runs perfect, the problem is when I add a new 
instruction, it doesn´t mather what instruction and where in my source 
code, it makes gcc to compile one of my C functions without his return 
instruction, causing my program to crash when I run it.

The two compiled versions of my C function are attached to this post.

Here are the last four changing instructions, before and after adding an 
arbitrary instruction:

before:
0x200007c4 <RCC_ADCCLKConfig+56>: add.w r7, r7, #20  ; 0x14
0x200007c8 <RCC_ADCCLKConfig+60>: mov   sp, r7
0x200007ca <RCC_ADCCLKConfig+62>: pop   {r7}
0x200007cc <RCC_ADCCLKConfig+64>: bx    lr

after:
0x200007d4 <RCC_ADCCLKConfig+56>: add.w r9, r7, #336860180  ;0x14141414
0x200007d8 <RCC_ADCCLKConfig+60>: lsls  r0, r0, #0
0x200007da <RCC_ADCCLKConfig+62>: lsls  r4, r3, #0
0x200007dc <RCC_ADCCLKConfig+64>: lsls  r1, r3, #0


I think the compiler does something wrong, but I don´t know how adding 
an arbitrary instruction causes gcc suddenly to compile wrong a 
function. If someone knows where can be the problem I will be very 
gratefull.

Thanks

Attached files:

GCC_code.txt (2.76 KB)

Reply Quote Edit Move Thread sperren Anmeldepflicht aktivieren Anpinnen Delete topic Thread mit anderem zusammenführen Report post

Clifford S. (clifford) 2009-04-01 13:40 #1214971

Rate this post

•	▲ useful ▼ not useful

From the address, it looks as if this code is located in RAM. And that 
your code has stomped all over it!

Generate assembler from the compiler not the debugger. That will show 
the code the compiler generated, not the state of the RAM after you 
corrupted it. The debugger disassembles whatever is in the memory - good 
or bad.


> I think the compiler does something wrong

Occam's razor should be applied here. From the gamut of possibilities, 
you are selecting the least likely. The GNU compiler is used 
successfully for many millions of lines of code throughout the world. It 
is unlikely that you have found a bug that easily that no one else has 
spotted, and more likely that your code is flawed.


If you cannot see it, your best bet is to post the code, but you should 
be looking at buffer overruns, invalid pointers, and modification of 
const data through non-const pointers for example.

Clifford

Reply Quote Edit Delete Report post

Ferdinand Fahlbusch Guest 2009-04-01 21:23 #1215508

Rate this post

•	▲ useful ▼ not useful

Thank you.

I finally found the problem, I knew that my program fitted into yhe RAM, 
but the stack is in RAM too, and colided with my code.

Reply Quote Edit Delete Report post

Clifford S. (clifford) 2009-04-01 22:19 #1215591

Rate this post

•	▲ useful ▼ not useful

Ferdinand Fahlbusch wrote:
> Thank you.
>
> I finally found the problem, I knew that my program fitted into yhe RAM,
> but the stack is in RAM too, and colided with my code.

... Ah, yes. Stack overflow, I should have mentioned that too. Stacks 
grow downward in memory (toward the code in this case).

Reply Quote Edit Delete Report post