RFC 7049 conformance issues

[maxgribov]

Recently tried wrap your lib in objective - c (for iOS/MAC OS client - server exchange purposes) and bump into couple issues conformance with RFC 7049 standard.

First: wrong byte order after encoding
RFC 7049, Appendix A

       //...
        int value = 1000000;
        cbor_item_t *cborItem = cbor_build_uint32(value);

        unsigned char *buffer;
        size_t buffer_size,
        length = cbor_serialize_alloc(cborItem, &buffer, &buffer_size);

        NSData *bufferData = [NSData dataWithBytes:(const void *)buffer length:sizeof(unsigned char)*length];
        //...

Result binary data (in ascii hex): 0x1a40420f00
Must be (according to RFC 7049): 0x1a000f4240

_All bytes after first byte in reverse order._

Second: negative values of all wides of integers encoding incorrect. For example, if I even reorder bytes in correct way - result not confirms to RFC 7049:

       //...
        short value = -1000;
        cbor_item_t *cborItem = cbor_build_uint16(value);
        cbor_mark_negint(cborItem);

        unsigned char *buffer;
        size_t buffer_size,
        length = cbor_serialize_alloc(cborItem, &buffer, &buffer_size);

        NSData *bufferData = [NSData dataWithBytes:(const void *)buffer length:sizeof(unsigned char)*length];
        //...

Result binary data (in ascii hex): 0x39fc18
Must be (according to RFC 7049): 0x3903e7

Third: floats encoding into something absolutely crazy. But any float encoded before with cbor_build_float4 can't be decoded with cbor_float_get_float4... It always returns 0.0

       //...
        case CBOR_TYPE_FLOAT_CTRL:

            switch (cbor_float_get_width(itemFromData)) {
                case CBOR_FLOAT_0:{

                    if (cbor_ctrl_is_bool(itemFromData)) {

                        resultNumber = [NSNumber numberWithBool:cbor_ctrl_value(itemFromData)];

                    }

                }
                    break;
                case CBOR_FLOAT_16:
                    resultNumber = [NSNumber numberWithFloat:cbor_float_get_float2(itemFromData)];
                    break;
                case CBOR_FLOAT_32:
                    resultNumber = [NSNumber numberWithFloat:cbor_float_get_float4(itemFromData)];
                    break;
                case CBOR_FLOAT_64:
                    resultNumber = [NSNumber numberWithDouble:cbor_float_get_float8(itemFromData)];
                    break;

            }

            break;
        //...

There is chance, that I am using your lib in some wrong way, but I quit trying understand this... Maybe you can help me?

[ekroon]

Maybe something went wrong with the compilation, you could try #8

[dgrigsby]

I have also seen the "First" issue described above for input: 100000.0
Expected: 0xFA47C35000 (per RFC 7049)
Observed: 0xFA0050C347

[dgrigsby]

@ekroon it's indeed possible this is related to #8.
I am currently running on the Mac XCode iOS simulator.
Highlighted line seems to be not working as expected:
size_t _cbor_encode_uint32(uint32_t value, unsigned char *buffer, size_t buffer_size, uint8_t offset)
{
if (buffer_size >= 5) {
buffer[0] = 0x1A + offset;

#ifdef HAVE_ENDIAN_H
*(uint32_t *) &buffer[1] = htobe32(value);
#else
#ifdef IS_BIG_ENDIAN
====> *(uint32_t *) &buffer[1] = value;
#else
buffer[1] = value >> 24;
buffer[2] = value >> 16;
buffer[3] = value >> 8;
buffer[4] = value;
#endif
#endif

    return 5;
} else
    return 0;

}

[dgrigsby]

Yes, I'm on 0.3.1 and IS_BIG_ENDIAN is incorrectly true for the XCode Simulator.
#8 would take care of this if it were incorporated, and if you use CMake to generate your xcodeproj.

[dgrigsby]

Issue is likely all #ifdef IS_BIG_ENDIAN should be changed to #if IS_BIG_ENDIAN.

[PJK]

I believe this has been fixed by @dgrigsby through #10