printf() problem used in the mc9s12dj64 in the CW 5.7.0

[zhangtracker]

I define the start_up_server function as follow.

unsigned short start_up_server (void * p_device)

{

...

reg_value = 0x55;

printf("**reg_value = 0x%02x\n",reg_value);
...

}

Call the start_up_server in the main function.  The printf function can't work normally.

output:

**reg_value = 0x30f9

note: 30f9 is the address of the p_device.

why?

[zhangtracker]

I didn't include stdio.h.

After I include stdio.h, printf funtion works normally.

Why?

Thanks  Lundin and CrasyCat.

[CrasyCat]

Hello

  printf is a function with an open parameter list. Parameter passing scheme is different for such  

   functions, so the compiler needs to know it should use a special parameter passing scheme.

CrasyCat

[zhangtracker]

Thank you for your reply.

I have try to define "reg_value" variable to volatile. But the output of printf function doesn't change.

You said:

>I have experienced strange behavior like this before because the optimizer wasn't behaving.

What dose  the optimizer need to be set?  I use none optimization. My complie command  line arguments is "-CPUHCS12 -Mb".

[Lundin]

That is quite a complicated topic. Some optimization is enabled by default, some is not. The Freescalers are better suitable for answering this question than me, though I would at least recommend -Onf which turns off "Create sub-functions with common code", since that one usually creates problems and does more harm than good.

[CrasyCat · Accepted solution]

Hello

Do you have included stdio.h on top of the source file where you are using printf?

Where is reg_value allocated (at which address)?

Can you try

printf("**reg_value = 0x%02x\n",(unsigned int) reg_value); instead?

CrasyCat

[zhangtracker]

reg_value is a variable not a pointer. It is defined as follow.

unsigned char reg_value;

And  I use the termio.c to make the output to UART0.

[Lundin]

Then indeed there is something wrong with the printf function, or perhaps with the optimizer settings. Have you tried looking at the disassembly or single-stepping through it? I have experienced strange behavior like this before because the optimizer wasn't behaving. You could try to set all variables to volatile and see if it works.

[Lundin]

What type is reg_value? If it is an allocated variable, did you try typecasting it to unsigned short? If it is a pointer, perhaps you forget the '*' operator?