Why two identical pointers do not compare equal with -O1? [duplicate] - c

This question already has answers here:
Can an equality comparison of unrelated pointers evaluate to true?
(4 answers)
Closed 3 years ago.
#include <stdio.h>
int main(void)
{
int a, b;
int *p = &a;
#ifdef __clang__
int *q = &b + 1;
#elif __GNUC__
int *q = &b - 1;
#endif
printf("%p %p %d\n", (void *)p, (void *)q, p == q);
}
C11 § 6.5.9 \ 6 says that
Two pointers compare equal if and only if both are null pointers, both are pointers to the same object (including a pointer to an object and a subobject at its beginning) or function, both are pointers to one past the last element of the same array object, or one is a pointer to one past the end of one array object and the other is a pointer to the start of a different array object that happens to immediately follow the first array object in the address space.
I have tested it four different ways:
Clang 9.0.1 with -01 option;
Clang 9.0.1 without any options;
GCC 9.2.0 with -01 option;
GCC 9.2.9 without any options.
The results are the following:
$ ./prog_clang
0x7ffebf0a65d4 0x7ffebf0a65d4 1
$ ./prog_clang_01
0x7ffd9931b9bc 0x7ffd9931b9bc 1
$ ./prog_gcc
0x7ffea055a980 0x7ffea055a980 1
$ ./prog_gcc_01
0x7fffd5fa5490 0x7fffd5fa5490 0
What is the correct behavior in this case?

What is the correct behavior in this case?
There is none. Comparing pointers to or one past the end of two completely unrelated objects is undefined behavior.
Per footnote 109 of the C11 standard (bolding is mine):
Two objects may be adjacent in memory because they are adjacent elements of a larger array or adjacent members of a structure with no padding between them, or because the implementation chose to place them so, even though they are unrelated. If prior invalid pointer operations (such as accesses outside array bounds) produced undefined behavior, subsequent comparisons also produce undefined behavior.

Two pointers compare equal if and only if both are null pointers,
they are not null
both are pointers to the same object (including a pointer to an object and a subobject at its beginning) or function
they do not point to the same object, nor a subobject, nor a function
both are pointers to one past the last element of the same array object,
they are not pointers to array elements.
or one is a pointer to one past the end of one array object and the other is a pointer to the start of a different array object that happens to immediately follow the first array object in the address space.
they are not pointers to array elements.
So, according to the standard, your pointers do not meet the requirements for comparing as equal, and should have never compared as equal.
Now, in your tests, in the first three cases, the pointers did in fact compare as equal. One can say that the compilers do not strictly adhere to the standard, because the standard says "if and only if", but as you have seen, clang and gcc without -O1 behave as if the standard said "if" without the "and only if" part. The compilers simply do not try to take extra measures to ensure that the "and only if" part is respected, so they allow the pointers to compare as equal, as a matter of pure coincidence, despite the fact that according to the standard, they shouldn't.
Since it was pure coincidence, in the last case the coincidence does not hold true anymore, due to a number of unknown reasons having to do with the compiler's implementation of optimizations. The compiler may have decided to reverse the order of the variables on the stack, or to put them farther away from each other, or who knows what.

Just to clarify a thing in other answers. Assume that you have two objects, a and b, a pointer p and make the pointer point to one of the objects with an offset, for instance p = &a + 1. Now p could just by coincidence have the same value as &b.
But - and this is important - this does NOT mean that p is pointing at b.
I take a comparison here. Maybe it's a bit strange but bear with me. Imagine that you see a road sign pointing towards a city. Now imagine that you decide to stand between the sign and the city. Does this mean that the sign is pointing at you? In this situation the answer could be quite philosophical, but it shows the point (haha). Even if one could argue for both yes and no, it's quite clear that the sign is not intended to point at you. In the C standard, they have chosen to interpret "pointing at" as "intentionally pointing at".

Related

Detecting array flattening trick with gcc

Some code flattens multidimensional arrays like this:
int array[10][10];
int* flattened_array = (int*)array;
for (int i = 0; i < 10*10; ++i)
flattened_array[i] = 42;
This is, as far as I know, undefined behaviour.
I am trying to detect cases like this with gcc sanitizers, however, neither -fsanitize=address nor -fsanitize=undefined work.
Is there a sanitizer option that I'm missing, or perhaps a different way to detect this at run time? Or maybe I am mistaken and the code is legal?
Edit: the sanitizers detect this access as an error:
array[0][11] = 42;
but do not detect this:
int* first_element = array[0];
first_element[11] = 42;
Furthermore, clang detects the first access statically, and gives out a warning
warning: array index 11 is past the end of the array (which contains 10 elements) [-Warray-bounds]
Edit: the above does not change if int in the declaration is replaced with char.
Edit: There are two potential sources of UB.
Accessing an object (of type int[10]) through an lvalue of an incompatible type (int).
Out-of-bounds access with a pointer of type int* and an index >=10 where the size of the underlying array is 10 (rather than 100).
Sanitizers don't seem to detect the first kind of violation. There's a debate whether this is a violation at all. After all, there's also an object of type int at the same address.
As for the second potential UB, the UB sanitizer does detect such access, but only if it is done directly via the 2D array itself and not via another variable that points to its first element, as shown above. I don't think the two accesses should differ in legality. They should be either both legal (and then ubsan has a false positive) or both illegal (and then ubsan has a false negative).
Edit: Appendix J2 says array[0][11] should be UB, even though it is only informative.
From a language lawyer point of view, this is generally seen as invalid code because the integers arrays are only of size 10 and the code does access past the declared array size. Yet it used to be a common idiom, and I know no compiler that would not accept it. Still with all real world compilers I know, the resulting program will have the expected behaviour.
After a second (in reality much more) reading of the C11 standard draft (n1570) the intent of the standard is still not clear. 6.2.5 Types § 20 says:
An array type describes a contiguously allocated nonempty set of objects with a
particular member object type, called the element type.
It makes clear that an array contains contiguously allocated objects. But IMHO is unclear about whether a contiguously allocated set of objects is an array.
If you answer no, then the shown code does invoke UB by accessing an array past it last element
But if you answer yes, then a set of 10 contiguous sets of 10 contiguous integers gives 100 contiguous integers and can be seen as an array of 100 integers. Then the shown code would be legal.
That latter acception seems to be common in the real word because it is consistent with dynamic array allocation: you allocate enough memory for a number of objects, and you can access that as if it had been declared as an array - and the allocation function ensures no alignment problem.
My conclusion so far is:
is it nice and clean code: certainly not and I would avoid it in production code
does it invokes UB: I really do not know and my personal opinion is probably no
Let us look at the code added in the edit:
array[0][11] = 42;
The compiler knows that array is declared as int[10][10]. So it knows that both indexes must be less than 10, and it can raise a warning.
int* first_element = array[0];
first_element[11] = 42;
first_element is declared as a mere pointer. Statically, the compiler has to assume that it can point inside an array of unknown size, so outside of a specific context, it is much harder to raise a warning. Of course for a human programmer it is evident that both way should be seen the same, but as a compiler is not required to emit any diagnostic for out of bounds array, efforts to detect them are left to the minimum and only trivial cases are detected.
In addition, when a compiler internally codes pointer arithmetics on common platforms, it just computes a memory address which is the original address and a byte offset. So it could emit the same code as:
char *addr = (char *) first_element; // (1)
addr += 11 * sizeof(int); // (2)
*((int *) addr) = 42; // (3)
(1) is legal because a pointer to any objet (here an int) can be converter to a pointer to char, which is required to point to the first byte of the representation of the object
(2) the trick here is that (char *) first_element is the same as (char *) array because the first byte of the 10*10 array is the first byte of the first int of the first row, and an single byte can only have one single address. As the size of array is 10 * 10 * sizeof(int), 11 * sizeof(int) is a valid offset in it.
(3) for the very same reason, (char *) &array[1][1] is addr because elements in an array are contiguous so their byte representation are also contiguous. And as a forth and back conversion between 2 types is legal and required to give back the original pointer, (int *) addr is (int*) ((char*) &array[1][1]). That means that dereferencing (int *) addr is legal and shall have the same effect as array[1][1] = 42.
This does not mean that first_element[11] does not involve UB. array[0] has a declared size which is 10. It just explains why all known compilers accepts it (in addition to not wanting to break legacy code).
The sanitizers are not especially good at catching out-of-bounds access unless the array in question is a complete object.
For example, they do not catch out-of-bounds access in this case:
struct {
int inner[10];
char tail[sizeof(int)];
} outer;
int* p = outer.inner;
p[10] = 42;
which is clearly illegal. But they do catch access to p[11].
Array flattening is not really different in spirit from this kind of access. Code generated by the compiler, and the way it is instrumented by sanitizers, should be pretty similar. So there's little hope that array flattening can be detected by these tools.
Multidimensional arrays are required to be contiguously allocated (C uses row-major). And there can't be any padding between elements of an array - though not stated explicitly in the standard, this can be inferred with array definition that says "contiguously allocated nonempty set of objects" and the definition of sizeof operator.
So the "flattening" should be legal.
Re. accessing array[0][11]: although, Annex J2 directly gives an example, what exactly is the violation in the normative isn't obvious. Nevertheless, it's still possible to make it legal an intermediate cast to char*:
*((int*)((char*)array + 11 * sizeof(int))) = 42;
(writing such code is obviously not advised ;)
The problem here is that there Standard describes as equivalent two operations, one of which clearly should be defined and one of which the Standard expressly says is not defined.
The cleanest way to resolve this, which seems to coincide with what clang and gcc already do, which is to say that applying [] operator to an array lvalue or non-l value does not cause it to decay, but instead looks up an element directly, yielding an lvalue if the array operand was an lvalue, and a non-l value otherwise.
Recognizing the use of [] with an array as being a distinct operator would clean up a number of corner cases in the semantics, including accessing an array within a structure returned by a function, register-qualified arrays, arrays of bitfields, etc. It would also make clear what the inner-array-subscript limitations are supposed to mean. Given foo[x][y], a compiler would be entitled to assume that y would be within the bounds of the inner array, but given *(foo[x]+y) it would not be entitled to make such an assumption.

Pointer to one before first element of array

It is said in C that when pointers refer to the same array or one element past the end of that array the arithmetics and comparisons are well defined. Then what about one before the first element of the array? Is it okay so long as I do not dereference it?
Given
int a[10], *p;
p = a;
(1) Is it legal to write --p?
(2) Is it legal to write p-1 in an expression?
(3) If (2) is okay, can I assert that p-1 < a?
There is some practical concern for this. Consider a reverse() function that reverses a C-string that ends with '\0'.
#include <stdio.h>
void reverse(char *p)
{
char *b, t;
b = p;
while (*p != '\0')
p++;
if (p == b) /* Do I really need */
return; /* these two lines? */
for (p--; b < p; b++, p--)
t = *b, *b = *p, *p = t;
}
int main(void)
{
char a[] = "Hello";
reverse(a);
printf("%s\n", a);
return 0;
}
Do I really need to do the check in the code?
Please share your ideas from language-lawyer/practical perspectives, and how you would cope with such situations.
(1) Is it legal to write --p?
It's "legal" as in the C syntax allows it, but it invokes undefined behavior. For the purpose of finding the relevant section in the standard, --p is equivalent to p = p - 1 (except p is only evaluated once). Then:
C17 6.5.6/8
If both the pointer
operand and the result point to elements of the same array object, or one past the last
element of the array object, the evaluation shall not produce an overflow; otherwise, the behavior is undefined.
The evaluation invokes undefined behavior, meaning it doesn't matter if you de-reference the pointer or not - you already invoked undefined behavior.
Furthermore:
C17 6.5.6/9:
When two pointers are subtracted, both shall point to elements of the same array object, or one past the last element of the array object;
If your code violates a "shall" in the ISO standard, it invokes undefined behavior.
(2) Is it legal to write p-1 in an expression?
Same as (1), undefined behavior.
As for examples of how this could cause problems in practice: imagine that the array is placed at the very beginning of a valid memory page. When you decrement outside that page, there could be a hardware exception or a pointer trap representation. This isn't a completely unlikely scenario for microcontrollers, particularly when they are using segmented memory maps.
The use of that kind of pointer arithmetic is bad coding practice, as it might lead to a significant bunch of hard to debug problems.
I only had to use this kind of thing once in more than 20 years. I was writing a call-back function, but I did not have access to the proper data. The calling function provided a pointer inside a proper array, and I needed the byte just before that pointer.
Considering that I had access to the entire source code, and I verified the behavior several times to prove that I get what I need, and I had it reviewed by other colleagues, I decided it is OK to let it go to production.
The proper solution would have been to change the caller function to return the proper pointer, but that was not feasible, time and money considering (that part of the software was licensed from a third-party).
So, a[-1] is possible, but should be used ONLY with very good care in very particular situations. Otherwise, there is no good reason to do that kind of self-hurting Voodoo ever.
Note: at a proper analysis, in my example, it is obvious that I did not access an element before the beginning of a proper array, but the element before a pointer, which was guaranteed to be inside of the same array.
Referring to the code provided:
it is NOT OK to use p[-1] with reverse(a);;
it is OK(-ish) to use it with reverse(a+1);, because you remain inside the array.

What harm would arise by pointer arithmetic beyond a valid memory range?

I followed the discussion on One-byte-off pointer still valid in C?.
The gist of that discussion, as far as I could gather, was that if you have:
char *p = malloc(4);
Then it is OK to get pointers up to p+4 by using pointer arithmetic. If you get a pointer by using p+5, then the behavior is undefined.
I can see why dereferencing p+5 could cause undefined behavior. But undefined behavior using just pointer arithmetic?
Why would the arithmetic operators + and - not be valid operations? I don’t see any harm by adding or subtracting a number from a pointer. After all, a pointer is a represented by a number that captures the address of an object.
Of course, I was not in the standardization committee :) I am not privy to the discussions they had before codifying the standard. I am just curious. Any insight will be useful.
The simplest answer is that it is conceivable that a machine traps integer overflow. If that were the case, then any pointer arithmetic which wasn't confined to a single storage region might cause overflow, which would cause a trap, disrupting execution of the program. C shouldn't be obliged to check for possible overflow before attempting pointer arithmetic, so the standard allows a C implementation on such a machine to just allow the trap to happen, even if chaos ensues.
Another case is an architecture where memory is segmented, so that a pointer consists of a segment address (with implicit trailing 0s) and an offset. Any given object must fit in a single segment, which means that valid pointer arithmetic can work only on the offset. Again, overflowing the offset in the course of pointer arithmetic might produce random results, and the C implementation is under no obligation to check for that.
Finally, there may well be optimizations which the compiler can produce on the assumption that all pointer arithmetic is valid. As a simple motivating case:
if (iter - 1 < object.end()) {...}
Here the test can be omitted because it must be true for any pointer iter whose value is a valid position in (or just after) object. The UB for invalid pointer arithmetic means that the compiler is not under any obligation to attempt to prove that iter is valid (although it might need to ensure that it is based on a pointer into object), so it can just drop the comparison and proceed to generate unconditional code. Some compilers may do this sort of thing, so watch out :)
Here, by the way, is the important difference between unspecified behaviour and undefined behaviour. Comparing two pointers (of the same type) with == is defined regardless of whether they are pointers into the same object. In particular, if a and b are two different objects of the same type, end_a is a pointer to one-past-the-end of a and begin_b is a pointer to b, then
end_a == begin_b
is unspecified; it will be 1 if and only if b happens to be just after a in memory, and otherwise 0. Since you can't normally rely on knowing that (unless a and b are array elements of the same array), the comparison is normally meaningless; but it is not undefined behaviour and the compiler needs to arrange for either 0 or 1 to be produced (and moreover, for the same comparison to consistently have the same value, since you can rely on objects not moving around in memory.)
One case I can think of where the result of a + or - might give unexpected results is in the case of overflow or underflow.
The question you refer to points out that for p = malloc(4) you can do p+4 for comparison. One thing this needs to guarantee is that p+4 will not overflow. It doesn't guarantee that p+5 wont overflow.
That is to say that the + or - themselves wont cause any problems, but there is a chance, however small, that they will return a value that is unsuitable for comparison.
Performing basic +/- arithmetic on a pointer will not cause a problem. The order of pointer values is sequential: &p[0] < &p[1] < ... &p[n] for a type n objects long. But pointer arithmetic outside this range is not defined. &p[-1] may be less or greater than &p[0].
int *p = malloc(80 * sizeof *p);
int *q = p + 1000;
printf("p:%p q:%p\n", p, q);
Dereferencing pointers outside their range or even inside the memory range, but unaligned is a problem.
printf("*p:%d\n", *p); // OK
printf("*p:%d\n", p[79]); // OK
printf("*p:%d\n", p[80]); // Bad, but &p[80] will be greater than &p[79]
printf("*p:%d\n", p[-1]); // Bad, order of p, p[-1] is not defined
printf("*p:%d\n", p[81]); // Bad, order of p[80], p[81] is not defined
char *r = (char*) p;
printf("*p:%d\n", *((int*) (r + 1)) ); // Bad
printf("*p:%d\n", *q); // Bad
Q: Why is p[81] undefined behavior?
A: Example: memory runs 0 to N-1. char *p has the value N-81. p[0] to p[79] is well defined. p[80] is also well defined. p[81] would need to the value N to be consistent, but that overflows so p[81] may have the value 0, N or who knows.
A couple of things here, the reason p+4 would be valid in such a case is because iteration to one past the last position is valid.
p+5 would not be a problem theoretically, but according to me the problem will be when you will try to dereference (p+5) or maybe you will try to overwrite that address.

C arrays and pointers: equivalent pointer to array element

Edit: The similar question asked before has not addressed some perspectives related to the issue.
In the ANSI C book by Kernighan and Ritchie, they say that the following are equivalent
a[i]
*(a+i)
I don't see how this can be true for elements that occupy more than one address space, e.g. structs.
Please explain? Edit: Thank you for all answers, but I don't quite understand it. It would seem I am suffering the same confusion as #CucumisSativus from his answer and comments to it.
Say sizeof(*a) is 3. If for some reason I wanted to access the middle byte of the first element in a, I had thought this is how I would do it: *(a+1).
Say the address of a is 10, and the sizeof each element is 20. And say we want to get the pointer to the second element. As I see it, we could do this: p = (10 + 20). I thought this would be equivalent to &a[1].
I'm having real trouble explaining what I don't understand!!
Pointer arithmetic is treated differently than regular integer arithmetic in C. Adding an integer i to a pointer p advances the memory address by i * sizeof(*p), i.e., by i times the size of the type being pointed to.
As a potentially interesting, but practically useless, sidenote: due to the definition of p[i] as *(p+i), the expression i[p] is also equivalent to the same…
The + operator is not meant to be the next element in the address space.
The increment is defined by the size of the data type involved. Thus, a + 1 will refer to the next int if a is a pointer to an integer, and will refer to the next struct if a is a pointer to a struct.
Lets say we have an Integer pointer and a Character pointer as follows
int *a;
char *b;
Now assume that a is stored at location 10000
and b is stored at location 20000
So by doing the following operations
*(a+1) : It will return the value contained in address 10004 i.e. (10000+4)
*(b+1) : It will return the value contained in address 20001 i.e. (20000+1)
Reason:
Pointer addition is different from regular arithmetic addition. If you add 1 to an integer pointer it advances the pointer to a location by the size of the integer. So in this case it advances the pointer to 10004. Since character is only 1 byte long it advances the pointer by 1 byte.
a+i is translated to something like this in pseudo code
*(a+i*sizeof(variableType))
I read more of K&R and found the missing pieces to the puzzle, allowing me to rationalise what's happening. (Whether I'm correct or not is open to debate!)
From what I gather, the following is simply invalid:
p = (10 + 20)
Given that, to quote: "Pointers and integers are not interchangeable. Zero is the only exception."
I interpret this to mean that *(a+i) is not performing straight forward integer arithmetic.
I believe this behaviour is determined by the existence of a pointer in the expression.
And to quote K&R again, and other answers provided to my question, the exact behaviour of arithmetic of arithmetic involving:
one pointer with
one or more integers
for (p + n) is defined like so: "n is scaled according to the size of the objects p points to, which is determined by the declaration of p".
I realised what caused me to be so confused: I have in the past been able to use printf() to output the actual address value of a pointer. I believe this is "undefined behaviour" which the compiler I was using happened to handle by treating the pointer as an integer.
Similarly, given the above two numbered bullet points, and the definition that uses them, I gather that adding two pointers together (of the same type) results in undefined behaviour, and more certainly, adding two pointers of different types.
To get the nth byte of an integer, see this question.
K&R later adds that the following are legal pointer operations:
Assigning one pointer to another (of the same type)
Adding or subtracting a pointer and an integer
Subtracting one pointer from another (to the same array)
Comparing pointers (to the same array)
Assigning / comparing to 0
All other pointer arithmetic is illegal.

Pointer arithmetic in c and array bounds

I was browsing through a webpage which had some c FAQ's, I found this statement made.
Similarly, if a has 10 elements and ip
points to a[3], you can't compute or
access ip + 10 or ip - 5. (There is
one special case: you can, in this
case, compute, but not access, a
pointer to the nonexistent element
just beyond the end of the array,
which in this case is &a[10].
I was confused by the statement
you can't compute ip + 10
I can understand accessing the element out of bounds is undefined, but computing!!!.
I wrote the following snippet which computes (let me know if this is what the website meant by computing) a pointer out-of-bounds.
#include <stdio.h>
int main()
{
int a[10], i;
int *p;
for (i = 0; i<10; i++)
a[i] = i;
p = &a[3];
printf("p = %p and p+10 = %p\n", p, p+10);
return 0;
}
$ ./a.out
p = 0xbfa53bbc and p+10 = 0xbfa53be4
We can see that p + 10 is pointing to 10 elements(40 bytes) past p. So what exactly does the statement made in the webpage mean. Did I mis-interpret something.
Even in K&R (A.7.7) this statement is made:
The result of the + operator is the
sum of the operands. A pointer to an
object in an array and a value of any
integral type may be added. ... The
sum is a pointer of the same type as
the original pointer, and points to
another object in the same array,
appropriately offset from the original
object. Thus if P is a pointer to an
object in an array, the expression P+1
is a pointer to the next object in the
array. If the sum pointer points
outside the bounds of the array,
except at the first location beyond
the high end, the result is
undefined.
What does being "undefined" mean. Does this mean the sum will be undefined, or does it only mean when we dereference it the behavior is undefined. Is the operation undefined even when we do not dereference it and just calculate the pointer to element out-of-bounds.
Undefined behavior means exactly that: absolutely anything could happen. It could succeed silently, it could fail silently, it could crash your program, it could blue screen your OS, or it could erase your hard drive. Some of these are not very likely, but all of them are permissible behaviors as far as the C language standard is concerned.
In this particular case, yes, the C standard is saying that even computing the address of a pointer outside of valid array bounds, without dereferencing it, is undefined behavior. The reason it says this is that there are some arcane systems where doing such a calculation could result in a fault of some sort. For example, you might have an array at the very end of addressable memory, and constructing a pointer beyond that would cause an overflow in a special address register which generates a trap or fault. The C standard wants to permit this behavior in order to be as portable as possible.
In reality, though, you'll find that constructing such an invalid address without dereferencing it has well-defined behavior on the vast majority of systems you'll come across in common usage. Creating an invalid memory address will have no ill effects unless you attempt to dereference it. But of course, it's better to avoid creating those invalid addresses so that your code will work perfectly even on those arcane systems.
The web page wording is confusing, but technically correct. The C99 language specification (section 6.5.6) discusses additive expressions, including pointer arithmetic. Subitem 8 specifically states that computing a pointer one past the end of an array shall not cause an overflow, but beyond that the behavior is undefined.
In a more practical sense, C compilers will generally let you get away with it, but what you do with the resulting value is up to you. If you try to dereference the resulting pointer to a value, as K&R states, the behavior is undefined.
Undefined, in programming terms, means "Don't do that." Basically, it means the specification that defines how the language works does not define an appropriate behavior in that situation. As a result, theoretically anything can happen. Generally all that happens is you have a silent or noisy (segfault) bug in your program, but many programmers like to joke about other possible results from causing undefined behavior, like deleting all of your files.
The behaviour would be undefined in the following case
int a[3];
(a + 10) ; // this is UB too as you are computing &a[10]
*(a+10) = 10; // Ewwww!!!!

Resources