| Details | Message |
|---|
Read-Only Author
Lars Hehl Posted
11-Jul-2005 12:23 GMT Toolset
C51 |  array index type of signed int Lars Hehl Hi list,
i'm working with c51v750a. My problem:
int array[ 0] _at_ 0x7600; //xdata
unsigned char index = 0; //xdata
printf("%d", &array[ index]);
printf("%d", &array[(unsigned char)(index - 1)]); //what is wrong?
// Result:
// 30208 --> 0x7600 (correct)
// 30206 --> 0x75FE == array[ -1] incorrect, // expected array[255] @ 0x77FE
I've read in Compiler's changelog [v7.50] that there was a fix for array with signed int index.
I don't need a workaround, but this seems like a compiler-bug. Or is my "logic" incorrect?
Thanks in advance,
Lars |
|
Read-Only Author
Mark Hickman Posted
11-Jul-2005 21:42 GMT Toolset
C51 | RE: array index type of signed int Mark Hickman Hi,
Not realy an answer but a question, doesn't the line 'array[0] _at 0x7600' generate a compiler error (invalid dimension size)? and should instead be 'array[0x0100] _at 0x7600'
But as regards your question, what you have written is what i would expect and to me suggests that keil are treating the index as signed even though you have declared/cast it as unsigned.
Mark. |
|
Read-Only Author
Andy Neil Posted
11-Jul-2005 22:24 GMT Toolset
C51 | RE: array index type of signed int Andy Neil What exactly are you trying to achieve here?
Is this some competition to use as many bizarre tricks in one go as possible?!
* Arrays of size zero
* negative array indexes (ie, attempting to access elements before the start of the array)
* using a negative index, but then casting it to signed and relying on it coming out as positive 255!
If you want to access X:0x76FF, why don't you just do it?
Or, if you want an index of 255, why don't you just do it - instead of obfuscating it with a cast of -1 ?!
I think the compiler might feel entitled to behave a little oddly in the face of such tactics! |
|
Read-Only Author
Lars Hehl Posted
11-Jul-2005 23:50 GMT Toolset
C51 | RE: array index type of signed int Lars Hehl oops, sorry,
there is an error in the sample code:
the first line has to be
unsigned int array[ 256] _at_ 0x7600
The code in my first post is only a example to show the problem. I've run into the problem in a more complex way. Then i tried this code above to check what' wrong. A simple typecast of index should generate a correct result. But the compiler didn't get it. I know, this code is far away from being 'smart', but it should work correct. Unfortunately, it doesn't. Are there other known problems with typecasting? I'm afraid, there are some other typecasts in my sourcecode. |
|
Read-Only Author
Andrew Neil Posted
12-Jul-2005 00:57 GMT Toolset
C51 | RE: array index type of signed int Andrew Neil "Are there other known problems with typecasting? I'm afraid, there are some other typecasts in my sourcecode."
But why are you casting -1 when what you really want is 255?
Why can't you just write 255?! |
|
Read-Only Author
Lars Hehl Posted
12-Jul-2005 07:44 GMT Toolset
C51 | RE: array index type of signed int Lars Hehl I want to create a circular buffer with size of 256 (0-255)integers. The output is LIFO. In my case, the index-counter is decremented. The type of the index counter is unsigned char (0-255). There is no problem when i first decrement the index and then read/write the array. But when i put the result of the calculation (index - 1)
array[(unsigned char)( index - 1)] //THIS IS AN EXAMPLE
the result of the index-counter is -1, for index = 0. The result without typecast is correct ( -1, int), but when i put a typecast before, it has to be 255!! I know how to avoid this problem, but in my opinion this is a bug in the compiler. |
|
Read-Only Author
A.W. Neil Posted
12-Jul-2005 09:20 GMT Toolset
C51 | RE: array index type of signed int A.W. Neil "The output is LIFO."
Surely, that's a stack - not a circular buffer? |
|
Read-Only Author
Lars Hehl Posted
12-Jul-2005 09:31 GMT Toolset
C51 | RE: array index type of signed int Lars Hehl "Surely, that's a stack - not a circular buffer?"
It's a mixture of both, but please, don't misinterprete the use of this code. To get the problem, you don't need to know what is the use of the code. I don't want to discuss about storage technics. |
|
Read-Only Author
Stefan Duncanson Posted
12-Jul-2005 11:06 GMT Toolset
C51 | RE: array index type of signed int Stefan Duncanson There's definitely something odd going on here. Take a look at this:
int xdata array[256] _at_ 0x7600; //xdata
unsigned char index = 0;
unsigned char idx;
idx=(unsigned char)(index - 1);
printf("%p\n", &array[(unsigned char)(index - 1)]); //x:75FE
printf("%p\n", &array[idx]); //x:75FE
printf("%p\n", array+(unsigned char)(index - 1)); //x:77FE
printf("%p\n", array+idx); //x:75FE
|
|
Read-Only Author
Stefan Duncanson Posted
12-Jul-2005 11:09 GMT Toolset
C51 | RE: array index type of signed int Stefan Duncanson Sorry, ingore my previous post. It's full of typos. This is hopefully correct:
int xdata array[256] _at_ 0x7600; //xdata
unsigned char index = 0;
unsigned char idx;
idx=(unsigned char)(index - 1);
printf("%p\n", &array[(unsigned char)(index - 1)]); //x:75FE
printf("%p\n", &array[idx]); //x:77FE
printf("%p\n", array+(unsigned char)(index - 1)); //x:75FE
printf("%p\n", array+idx); //x:77FE
|
|
Read-Only Author
Lars Hehl Posted
12-Jul-2005 11:43 GMT Toolset
C51 | RE: array index type of signed int Lars Hehl Stefan,
did you compile with c51v750a ? BTW, the software-versions which are displayed in the uV2 are:
C-Compiler C51.exe v7.50
Assembler A51.exe v7.10
Linker BL51.exe v5.12
Librarian LIB51.exe v4.24
I'm sure i've updated to C51v750a.
I will compile the example code with an older version and post the result here. |
|
Read-Only Author
Stefan Duncanson Posted
12-Jul-2005 12:26 GMT Toolset
C51 | RE: array index type of signed int Stefan Duncanson "did you compile with c51v750a ?"
No, C51 v7.01. |
|
Read-Only Author
Lars Hehl Posted
12-Jul-2005 12:42 GMT Toolset
C51 | RE: array index type of signed int Lars Hehl "No, C51 v7.01."
ok, then i don't need test it with my older version V7.04
I will contact Keil for informations, thanks. |
|
Read-Only Author
Stefan Duncanson Posted
12-Jul-2005 13:58 GMT Toolset
C51 | RE: array index type of signed int Stefan Duncanson "I will contact Keil for informations, thanks."
Please post the result when you get one. |
|
Read-Only Author
Lars Hehl Posted
13-Jul-2005 08:20 GMT Toolset
C51 | RE: array index type of signed int Lars Hehl Today i received a mail from Keil:
Mr. Schneebauer wrote:
"prinzipiell haben Sie Recht, daß es sich hier um einen Fehler handelt.
Normalerweise versucht der Compiler das '-1' in die Konstanten zu
verpacken um das Programm besser zu optimieren. Wir haben diesen Fehler
nun auch behoben. Da Änderungen an zentralen Funktionen gemacht wurden,
muß dieser Compiler noch ausführlich getestet werden bevor wir ihn
ausliefern können."
Basicly, what he writes is:
Yes, its a bug. The bug is fixed and the new compiler-version will be tested.
Stefan, thank you for verifying this problem. |
|
