Difference between revisions of "cpp/language/ub"

Revision as of 16:17, 1 November 2024

Renders the entire program meaningless if certain rules of the language are violated.

Explanation

The C++ standard precisely defines the observable behavior of every C++ program that does not fall into one of the following classes:

• ill-formed - the program has syntax errors or diagnosable semantic errors.

• A conforming C++ compiler is required to issue a diagnostic, even if it defines a language extension that assigns meaning to such code (such as with variable-length arrays).
• The text of the standard uses shall, shall not, and ill-formed to indicate these requirements.

• ill-formed, no diagnostic required - the program has semantic errors which may not be diagnosable in general case (e.g. violations of the ODR or other errors that are only detectable at link time).

• The behavior is undefined if such program is executed.

• implementation-defined behavior - the behavior of the program varies between implementations, and the conforming implementation must document the effects of each behavior.

• For example, the type of std::size_t or the number of bits in a byte, or the text of std::bad_alloc::what.
• A subset of implementation-defined behavior is locale-specific behavior, which depends on the implementation-supplied locale.

• unspecified behavior - the behavior of the program varies between implementations, and the conforming implementation is not required to document the effects of each behavior.

• For example, order of evaluation, whether identical string literals are distinct, the amount of array allocation overhead, etc.
• Each unspecified behavior results in one of a set of valid results.

#include <cassert>
#include <cstring>
 
void f()
{   
    int d1, d2;       // d1, d2 have erroneous values
    int e1 = d1;      // erroneous behavior
    int e2 = d1;      // erroneous behavior
    assert(e1 == e2); // holds
    assert(e1 == d1); // holds, erroneous behavior
    assert(e2 == d1); // holds, erroneous behavior
 
    std::memcpy(&d2, &d1, sizeof(int)); // no erroneous behavior, but
                                        // d2 has an erroneous value
 
    assert(e1 == d2); // holds, erroneous behavior
    assert(e2 == d2); // holds, erroneous behavior
}
 
unsigned char g(bool b)
{
    unsigned char c;     // c has erroneous value
    unsigned char d = c; // no erroneous behavior, but d has an erroneous value
    assert(c == d);      // holds, both integral promotions have erroneous behavior
    int e = d;           // erroneous behavior
    return b ? d : 0;    // erroneous behavior if b is true
}

• undefined behavior - there are no restrictions on the behavior of the program.

• Some examples of undefined behavior are data races, memory accesses outside of array bounds, signed integer overflow, null pointer dereference, more than one modifications of the same scalar in an expression without any intermediate sequence point(until C++11)that is unsequenced(since C++11), access to an object through a pointer of a different type, etc.
• Implementations are not required to diagnose undefined behavior (although many simple situations are diagnosed), and the compiled program is not required to do anything meaningful.

UB and optimization

Because correct C++ programs are free of undefined behavior, compilers may produce unexpected results when a program that actually has UB is compiled with optimization enabled:

For example,

Signed overflow

int foo(int x)
{
    return x + 1 > x; // either true or UB due to signed overflow
}

may be compiled as (demo)

foo(int):
        mov     eax, 1
        ret

Access out of bounds

int table[4] = {};
bool exists_in_table(int v)
{
    // return true in one of the first 4 iterations or UB due to out-of-bounds access
    for (int i = 0; i <= 4; i++)
        if (table[i] == v)
            return true;
    return false;
}

May be compiled as (demo)

exists_in_table(int):
        mov     eax, 1
        ret

Uninitialized scalar

std::size_t f(int x)
{
    std::size_t a;
    if (x) // either x nonzero or UB
        a = 42;
    return a;
}

May be compiled as (demo)

f(int):
        mov     eax, 42
        ret

#include <cstdio>
 
int main()
{
    bool p; // uninitialized local variable
    if (p)  // UB access to uninitialized scalar
        std::puts("p is true");
    if (!p) // UB access to uninitialized scalar
        std::puts("p is false");
}

p is true
p is false

Invalid scalar

int f()
{
    bool b = true;
    unsigned char* p = reinterpret_cast<unsigned char*>(&b);
    *p = 10;
    // reading from b is now UB
    return b == 0;
}

May be compiled as (demo)

f():
        mov     eax, 11
        ret

Null pointer dereference

The examples demonstrate reading from the result of dereferencing a null pointer.

int foo(int* p)
{
    int x = *p;
    if (!p)
        return x; // Either UB above or this branch is never taken
    else
        return 0;
}
 
int bar()
{
    int* p = nullptr;
    return *p; // Unconditional UB
}

may be compiled as (demo)

foo(int*):
        xor     eax, eax
        ret
bar():
        ret

Access to pointer passed to std::realloc

#include <cstdlib>
#include <iostream>
 
int main()
{
    int* p = (int*)std::malloc(sizeof(int));
    int* q = (int*)std::realloc(p, sizeof(int));
    *p = 1; // UB access to a pointer that was passed to realloc
    *q = 2;
    if (p == q) // UB access to a pointer that was passed to realloc
        std::cout << *p << *q << '\n';
}

12

Infinite loop without side-effects

#include <iostream>
 
bool fermat()
{
    const int max_value = 1000;
 
    // Non-trivial infinite loop with no side effects is UB
    for (int a = 1, b = 1, c = 1; true; )
    {
        if (((a * a * a) == ((b * b * b) + (c * c * c))))
            return true; // disproved :()
        a++;
        if (a > max_value)
        {
            a = 1;
            b++;
        }
        if (b > max_value)
        {
            b = 1;
            c++;
        }
        if (c > max_value)
            c = 1;
    }
 
    return false; // not disproved
}
 
int main()
{
    std::cout << "Fermat's Last Theorem ";
    fermat()
        ? std::cout << "has been disproved!\n"
        : std::cout << "has not been disproved.\n";
}

Fermat's Last Theorem has been disproved!

Ill-formed with diagnostic message

Note that compilers are permitted to extend the language in ways that give meaning to ill-formed programs. The only thing C++ standard requires in such cases is a diagnostic message (compiler warning), unless the program was "ill-formed no diagnostic required".

For example, unless language extensions are disabled via --pedantic-errors, GCC will compile the following example with only a warning even though it appears in the C++ standard as an example of an "error" (see also GCC Bugzilla #55783)

#include <iostream>
 
// Example tweak, do not use constant
double a{1.0};
 
// C++23 standard, §9.4.5 List-initialization [dcl.init.list], Example #6:
struct S
{
    // no initializer-list constructors
    S(int, double, double); // #1
    S();                    // #2
    // ...
};
 
S s1 = {1, 2, 3.0}; // OK, invoke #1
S s2{a, 2, 3}; // error: narrowing
S s3{}; // OK, invoke #2
// — end example]
 
S::S(int, double, double) {}
S::S() {}
 
int main()
{
    std::cout << "All checks have passed.\n";
}

main.cpp:17:6: error: type 'double' cannot be narrowed to 'int' in initializer ⮠
list [-Wc++11-narrowing]
S s2{a, 2, 3}; // error: narrowing
     ^
main.cpp:17:6: note: insert an explicit cast to silence this issue
S s2{a, 2, 3}; // error: narrowing
     ^
     static_cast<int>( )
1 error generated.

References

See also

External links