is this bit field of size one actually overflowing when assigning 1?

Question

I have written some code with bit fields that I thought should work, but it seems like GCC disagrees. Did I miss something or did I actually find a bug in GCC?

After simplifying my code, the testcase is quite simple. I'm assigning the integer literal 1 to a bitfield that has a size of one bit:

typedef struct bitfield
{
    int bit : 1;
} bitfield;

bitfield test()
{
    bitfield field = {1};
    return field;
}

If I compile this with GCC 6.2.1 (same with 5.4.0), I get the following warning (with -pedantic):

gcc -fPIC test.c -pedantic -shared
test.c: In function ‘test’:
test.c:8:23: warning: overflow in implicit constant conversion [-Woverflow]
     bitfield field = {1};
                       ^

The strange thing is: When I replace -pedantic with -Woverflow, the warning disappears.

I don't get any warnings with clang.

Answer 1

Use unsigned int for this bit field. A 1-bit signed number can only hold 0 and -1.

typedef struct bitfield
{
    unsigned int bit : 1;
} bitfield;

Answer 2

is this bit field of size one actually overflowing when assigning 1?

Maybe. It is implementation defined.

---

This is a place in C where int and signed int/signed may differ.

unsigned int bit : 1; can hold the value 0 or 1.

Assuming 2's complement machine ...

signed int bit : 1; can hold the value 0 or -1
signed bit : 1; can hold the value 0 or -1

int bit : 1; can hold the value 0 or 1 or 0 or -1

Each of the comma-separated multisets (int, signed, or signed int) designates the same type, except that for bitfields, it is implementation-defined whether the specifier int designates the same type as signed int or the same type as unsigned int. C11 dr §6.7.2 5

---

Using unsigned removes ambiguities.