Will it erroneous ??

Maybe.

Your pointer is to Xdata, but the array is in the defualt space. For large RAM model it will work.

Maybe.

Your pointer is to Xdata, but the array is in the defualt space. For large RAM model it will work.

Well, my processor if basically a dual core consist of C51 MPU code and DSP Compute Engine core. But i think the problem is independent of processor, and somehow related to the code generated by Keil. There are 2 WD Timers. One is hardware and one is 8051 s/w. I disabled both, so no issue with rebooting by WD timer reset.

The strange behavior i observed in above code snippet is, when i call the test() function as below, system is misbehaving:

void main(void) { unsigned char i;

   while(1) {
      for(i=0; i<8; i++) {
         magicString[i] += 1;
      }
      test(&magicString);   // Passing address
   }
}

void test(unsigned char const xdata* buffer) {...}

But if i called the test() as below in above code, the system works fine:

test((unsigned char xdata *)&magicString); //Casting
test(magicString); // W/o casting

Its hard time for me to understand whats wrong with passing address of magicString i.e. test(&magicString) ?

Can anybody explain me why the last 2 test() work fine and not the first.

Its hard time for me to understand whats wrong with passing address of magicString i.e. test(&magicString) ?

"&magicString" is not the address of the first element of the array magicString[]. That's where the error is.

What you actually want is "&magicString[0]" or just "magicString", which both are a pointer to the first element of magicString.

""&magicString" is not the address of the first element of the array magicString[]."

If it isn't, then what is it?

The types that they point to are different but the address should be same. That's my understanding anyhow. (I haven't got a copy of the C standard before you ask though!)

void test(unsigned char const xdata* buffer) { unsigned char j;
for (j=8; j>0; j--) { send(*buffer++); // another function }
}

Should you be able to increment a const pointer passed as a parameter to a function?

"buffer" is not const, given that declaration. It points to a const char, but is not itself const.

const is left-associative. (If there's nothing to its left, it will modify the thing to its right.)

Another way to look at it is that anything to the left of the * describes the type pointed to; anything to the right describes the pointer itself.

The types that they point to are different but the address should be same.

Yes, that's right. I had hoped the person who thought otherwise would realise this in the process of trying to answer the question.