How to interpret binary data as an integer?

Question

The codebase at work contains some code that looks roughly like this:

#define DATA_LENGTH 64

u_int32 SmartKey::SerialNumber()
{
    unsigned char data[DATA_LENGTH];
    // ... initialized data buffer
    return *(u_int32*)data;
}

This code works correctly, but GCC gives the following warning:

warning: dereferencing pointer ‘serialNumber’ does break strict-aliasing rules

Can someone explain this warning? Is this code potentially dangerous? How can it be improved?

Update
With thanks to James McNellis' answer I came up with the following utility function:

template<class T, class Data>
T BinaryCast(const Data & inData)
{
    T ret;
    std::copy(&inData[0], &inData[0] + sizeof(ret), reinterpret_cast<char*>(&ret));
    return ret;
}

u_int32 SmartKey::SerialNumber()
{
    unsigned char data[DATA_LENGTH];
    // ... initialized data buffer
    return BinaryCast<u_int32>(data);
}

Feel free to suggest improvements!

Probably has to do with pointer being cast from unsigned char * to u_int32 * but been a long time since I did C++. If so, since DATA_LENGTH is an exact multiple of 32 there should not be an issue. — Eric J.
– Eric J., Commented May 27, 2010 at 15:27

Community · Accepted Answer · 2017-05-23 12:33:45Z

11

The warning is because you are violating the strict aliasing rule.

One way to do it correctly would be to copy the bytes from the data buffer into a u_int32 object and return that object:

unsigned char data[DATA_LENGTH];
// ... initialized data buffer

u_int32 i;
assert(sizeof (i) <= DATA_LENGTH);
std::copy(&data[0], &data[0] + sizeof (i), reinterpret_cast<char*>(&i));
return i;

This solution works because in C++ it is permitted to access any type of object as an array of char.

(std::copy() is in <algorithm>)

edited May 23, 2017 at 12:33

CommunityBot

11 silver badge

answered May 27, 2010 at 15:26

James McNellis

357k76 gold badges934 silver badges983 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

PeterK Over a year ago

Exactly what i wanted to post. Although my solution used memcpy.

user347594 Over a year ago

Even if you are not using STL, at least use reinterpret_cast.

user168715 Over a year ago

One subtlety that should be mentioned whenever talking about (de)serializing data in this way is that of endianness. It's less of a concern now that Apple uses Intel chips, but should be kept in mind lest your data becomes severely corrupted.

AnT stands with Russia · Accepted Answer · 2010-05-27 15:32:43Z

In C and C++ languages reinterpreting memory occupied by object of one type as an object of another type is illegal - it leads to undefined behavior. Some compilers use this rule to perform aggressive aliasing-related optimizations. As a consequence, your code might not work as expected, if you perform the aforementioned reinterpretation.

In C/C++ it is OK to reinterpret any object as an array of char, but it is not OK to take a standalone array of char and reinterpret is as an object of some other type. This is what your code is doing.

In addition to aliasing problems, you have to keep in mind that a standalone automatic char array is not guaranteed to be aligned properly to be read as an u_int32 value.

The proper way to do what the above code is trying to do is to copy the source array into an intermediate u_int32 value using memcpy

u_int32 SmartKey::SerialNumber()
{
    unsigned char data[DATA_LENGTH];
    u_int32 u;
    // ...
    memcpy(&u, data, sizeof u);
    return u;
}

Of course, you have to be sure that the endianness of the data is the same as the endianness of the u_int32 objects on your platform.

Vagrant · Accepted Answer · 2010-05-27 15:53:34Z

0

I think the problem is actually somewhere in your elided code to initialize the data[] structure. I don't think it has anything to do with your cast, which is fine.

answered May 27, 2010 at 15:53

Vagrant

1,71612 silver badges19 bronze badges

Comments

pronebird · Accepted Answer · 2010-05-27 15:53:02Z

-1

Not sure, but I think you can do like that:

return (u_int32)&data;

answered May 27, 2010 at 15:53

pronebird

12.3k5 gold badges57 silver badges84 bronze badges

1 Comment

AnT stands with Russia Over a year ago

That will convert the address of data to u_int32. The OP wants to read the actual content stored in first elements of data. Not even remotely the same thing.

Collectives™ on Stack Overflow

How to interpret binary data as an integer?

4 Answers 4

3 Comments

Comments

Comments

1 Comment

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

4 Answers 4

3 Comments

Comments

Comments

1 Comment

Your Answer

Sign up or log in

Post as a guest

Linked

Related