Hello!
I have uVision that compiles fine with the C51 v7.03 compiler and the related package, but not complete with the 8.06. I used two different Keil installations. All files are in the same folder.
In the 8.06 I get linker errors like "object does not fit in to pdata page" and "0080H". This looks like the compiler was thinking the PDATA was only 128 bytes, but it is set to 256 bytes in the startup.a51. Any clue what's different in the newer Keil package?
Also there is a warning in 8.06 (which does not show in 7.03) "converting non-pointer to pointer" on this
ptr_xdata = sPtr_obj->Adresse;
while the vars are set like this:
uchar uc_set_obj( uchar pdata *ptr_Set) { uchar i; uchar xdata *ptr_xdata; struct stOBJADR code *sPtr_obj; sPtr_obj=&Obj[*ptr_Set]; . . . ptr_xdata = sPtr_obj->Adresse; }
The struct stOBJADR has a member "uint Adresse;"
I can see no wrong use of the pointers. I just want to be sure that the warning does not affect the code to not work correctly.
As you say, ptr_xdata is a pointer, and sPtr_obj->Adresse isn't - so the warning is perfectly correct!
As usual, an explicit cast should both stop the warning and make your intention clear...
"I just want to be sure that the warning does not affect the code to not work correctly"
There is always a risk in playing fast-and-loose with pointers like this!
I think you should be OK with the XDATA-Specific pointer...
Hello Andy! Thanks for your help.
You are wrong. "sPtr_obj->Adresse" is a struct pointer and gives me a number, an address that "ptr_xdata" is set to.
And for the x:th time: you can not absolutely know about pointers and how it is used if you think that the address of a pointer is less important (or more or less irrelevant) than the address of the variable it points to.
Basically, you are ignoring arithmetic on pointers, which is something that is very commonly done.
When you do choose to see a pointer's address, a pointer's value and the value it points to as three separate things, your life will be a lot easier.
char buf[100]; char *p = buf+1; char **pp = &p; *buf = 'H'; *p++ = 'e'; p[0] = 'l'; p[1] = 'l'; buf[4] = 'o'; *(buf+5) = '\0'; *pp = strchr(buf,'\0'); printf("Number of characters in string is %ld\n",(long)(p-buf));
So when reading this value I actually do the same as setting the pointer value directly to the address of the variable And for that, you should tell your friendly compiler that the number you have really is an address - the address of an object of the same type that your left-hand-object (the pointer) is expected to point at.
You are assuming that a pointer is a number and because of this, the compiler should always accept the assign of any number to any pointer without getting upset.
Note that pound and dollar are two currencies. The amount of money are always numeric. But you can't simply assign a dollar value to a pound purchase. If you try to do that, the shop will issue a format conversion warning. The conversion from one currency to a different currency requres a type cast. For currencies, this is normally done by checking the current sell price for dollar and the current buy price for pound. In the case of converting from one currency to another, there is a change to the numeric value. The same thing happens when performing an assign from an integer to a floating point number, or the reverse.
In some special cases, there is no actual conversion performed. But the cast should be there anyway. Not because a change is needed to the numeric value, but because a change is needed to the data type, i.e. to tell the compiler that the object on the left and the object on the right are both of the same data type (or at least of compatible data types) - even if a quick look says something else.
Adding a type cast when assigning a number to a pointer tells the compiler that the numeric source value is a proper value for storage in the pointer. In this case, that the number really is a pointer value and suitable for the pointer you want to assign to.
This is no different from the following:
int a = 5; unsigned b = 5; if (a == b) ...
A lot of compilers will notice that a is a signed integer, and b is an unsigned integer, and will issue a warning about a comparison between signed and unsigned data. The warning is there because there is a reasonable chance that the signed/unsigned mismatch is caused by an invalid assumption somewhere.
When you add a typecast, the compiler will check if it is technically possible to treat the right-hand value as compatible with the left-hand value. If you don't write an explicit typecast, the compiler will check it's available automagic data type rewrite rules (as specified by the standard) to see if it somehow (with or without data loss) can convert the right-hand value into something suitable for the left-hand value. However, unless the compiler is really, really happy about any conversion rules it knows about it will complain by issuing a warning. Some compilers have a larger set of warnings than others. But when they do warn, you should care, because that means you are living on the edge.
In your case, a proper cast will tell your compiler that you have turned off the power and removed the fuse before doing electrical work in your house. The cast will mute the warning since you are giving the compiler a written contract where you, the developer, dearly promises that the value in your uint really, really do represent a good pointer to a variable of the correct data type.
int a = 5; unsigned b = 5; a = b;
Could generate a warning, because some legal unsigned values won't fit in a signed int
Similarly,
int a = 5; unsigned b = 5; b = a;
Could generate a warning, because some legal signed values won't fit into an unsigned int.
In both cases, you can tell the compiler that it's not an accidental oversight by giving an explicit cast.
Exactly the same applies when assigning a non-pointer to a pointer, as has already been explained.
"In my understanding and I guess in everyone else too, this means that I was trying to convert something that's not a pointer (a normal variable, for exmple) to a pointer."
Yes correct: The pointer is ptr_xdata; The non-pointer is sPtr_obj->Adresse.
"Since ptr_xdata is declared as a pointer, it will still be a pointer after its value has changed."
Correct - but the warning is not about that: Read it again, the warning is about converting a non-pointer to pointer - that is, the thing being converted is the non-pointer
ptr_xdata = (sPtr_obj->Adresse); ^ ^ | | | | This is a pointer This is not a pointer (it's a uint)
So the above assignment is from a non-pointer (on the right-hand side) to a pointer (on the left-hand side)
And so the warning is correct:
"converting [from] non-pointer to pointer"
You don't seem to get it. I don't understand it, because it's confusing.
In this case, it is actually the reverse. It is confusing because you don't understand it.
A number of million students have thought that pointers are confusing. However, there are quite a lot of programmers in this world that do know about pointers now. And they can testify that as soon as they really did understand pointers, they no longer see them as confusing.
It's just a question of getting the correct internal picture of the pointer concept. Before you have a working mental picture, they will be confusing. With the picture in place, you will think: gosh, that was easy.
The reason so many programs fails because of pointer errors, is not because pointers are hard to grasp but because they are powerful and a more powerful tool is inherently more dangerous. People makes mistakes. When they make mistakes with pointers, bad things happens. So a developer should always be extra careful when working with pointers.
Right now, we are trying to get you to change your internal view of pointers. A number of us believe that the terminology really is crucial to this task. If you see a pointer as a variable, with the normal properties of a variable (location, value, size, ...) then we believe you have a better chance to be able to switch your mental image of pointers, and be able to figure out their full potential. And their full potential really do require you to think about the address of the pointer, i.e. where it is stored. You really do have to understand why several steps of indirection is needed, or why programs must be able to modify the value of the pointer, and not just the value it points to.
I repeat for the xth time that I absolutely KNOW
And everybody else has to keep repeating, that you don't know anywhere as much as you think.
You're contradicting yourself all the time. One moment you claim you understand pointers perfectly, the next you admit you're confused by them.
You've stated painfully obvious nonsense like this, several times by now
this value is an uint value and I put this value to another pointer, also of type uint,
It shouldn't need saying yet another time, but apparently it does: No, that pointer is not of type uint. It's of type pointer-to-uint. A uint is not a pointer any more than a car is a driver's license. Staying in that image, you got a warning because you tried to put the car in your inside pocket.
Everybody, including you, is allowed to be wrong once in a while. But that doesn't mean we'll seeingly tolerate your insisting on staying a confused fool forever.
"One one man stands against the world, the world is often right."
To the insane, the rest of the world is crazy and they are the ones who are sane.
"the struct [element?] was formerly initialised with an address of a variable"
I presume you're talking about sPtr_obj->Adresse - yes?
If this is supposed to hold the address of a variable, why was it not properly defined as a pointer in the first place?!
The only reason why the language allows conversion from integers to pointers is for systems programming (embedded or when you write hardware drivers), i.e. when you need to access a hardware device at a specific location or when you need to locate variables dynamically at absolute locations.
Taking the address of a variable and convert to an integer and later converting the integer back to a pointer is something a program should hardly ever need. Most required address manipulation can be performed on the pointer. If you implement a memory manager, you might like to play with / and % (or more probably with >> and &) to figure out memory page and offset within memory page.
You can lead a horse to water, but you can't make him drink.
"You can lead a horse to water, but you can't make him drink."
Similarly:
"You can lead a boy to college, but you can't make him think." - Elbert Hubbard.
"You can lead a boy to college, but you can't make him think." - Elbert Hubbard. "you can lead a boy to college, but you can not make him do his homework instead of posting "please send schematic and code" in a forum"
Erik
I presume you're talking about sPtr_obj->Adresse - yes? If this is supposed to hold the address of a variable, why was it not properly defined as a pointer in the first place?!
Because it does not really work to access a pointer address value inside a struct that is pointed to by a pointer. It should be something like
ptr_xdata = *(sPtr_obj->Adresse)
if Adresse wouldn't be an uint, but a pointer instead. We had severe problems in the past calming down the compiler when using such a way. The chance is 50:50 that it will work. Means, the compiler was unsure how to handle it.
Means, the compiler was unsure how to handle it.
That is a statement to print and put on the wall. The next time I get a compiler error, I will complain about insecure compilers. That is almost as funny as saying that the car stood still when the tree suddenly decided to run into it.
You can have any (!) level of pointer indirection (with or without any structs in the indirection chain) without any problems separating operations on the pointer value (the pointer) or on the referenced object. It all follows from the operator precedence rules.
"Because it does not really work to access a pointer address value (sic) inside a struct that is pointed to by a pointer"
Doesn't it?
If the struct element is correctly defined as a pointer, then there is no difference whether it is read via a direct access as
ptr_xdata = some_struct.some_pointer
or via a pointer as
ptr_xdata = struct_ptr->some_pointer
"We had severe problems in the past calming down the compiler when using such a way"
I very much suspect that all comes back to you not really understanding what's going on.
"chance is 50:50 that it will work"
The compiler will do exactly what you tell it to do - but, as mentioned earlier, if you cannot express what you want correctly, then the compiler cannot guess for you...
"the compiler was unsure how to handle it."
Again, the compiler does what you tell it - but, if you tell it rubbish, you will get rubbish as a result.
Hence the well-known programming term, GIGO = Garbage In; Garbage Out.
Because it does not really work to access a pointer address value inside a struct that is pointed to by a pointer.
It should work just fine as long as proper attention is paid to things like operator precedence. This functionality is well-defined and definitely not compiler-specific. If it does not work, then there is either a mistake in the C code or a bug in the compiler. The former is usually much more likely than the latter.
It should be something like
No. It shouldn't be something like X - this leads to bugs. The syntax for doing that is very clear and if you would not insist on ignoring how to work with pointers in C, you would probably know how to do it.
Have you read up on pointer-specific operators, especially the "Member access from pointer" operator, "->" ?
What is this supposed to do ?
1. Copy the value of the structure member Adresse to ptr_xdata ? 2. Copy the address of the structure member Adresse to ptr_xdata ? 3. Copy the contents of the memory address pointed to by the structure member Aresse to ptr_xdata ?
We had severe problems in the past calming down the compiler when using such a way.
The compiler should have absolutely no trouble parsing a trivial operation like that, provided it is fed with the correct code.
The chance is 50:50 that it will work.
Compilers don't roll dice. Provided that you didn't switch compilers somewhere in the middle, there should be no statistics involved. The code is either correct, or it is not.
View all questions in Keil forum