Is it legal to convert short to char?

Question

unsigned char HexLookup[513] = {
    "000102030405060708090a0b0c0d0e0f"
    "101112131415161718191a1b1c1d1e1f"
    "202122232425262728292a2b2c2d2e2f"
    "303132333435363738393a3b3c3d3e3f"
    "404142434445464748494a4b4c4d4e4f"
    "505152535455565758595a5b5c5d5e5f"
    "606162636465666768696a6b6c6d6e6f"
    "707172737475767778797a7b7c7d7e7f"
    "808182838485868788898a8b8c8d8e8f"
    "909192939495969798999a9b9c9d9e9f"
    "a0a1a2a3a4a5a6a7a8a9aaabacadaeaf"
    "b0b1b2b3b4b5b6b7b8b9babbbcbdbebf"
    "c0c1c2c3c4c5c6c7c8c9cacbcccdcecf"
    "d0d1d2d3d4d5d6d7d8d9dadbdcdddedf"
    "e0e1e2e3e4e5e6e7e8e9eaebecedeeef"
    "f0f1f2f3f4f5f6f7f8f9fafbfcfdfeff"
};

void to_hex(char* in_data, unsigned int in_data_size, char* out_data)
{
    unsigned short *pwHex =  (unsigned short*)HexLookup;

    for (unsigned int i = 0; i < in_data_size; i++)
    {
        *out_data = pwHex[*in_data];
        in_data++;
        out_data++;
    }
}

In the code above, what is the behavior of:

unsigned short *pwHex = (unsigned short*)HexLookup;
*out_data = pwHex[*in_data];

Beware of alignment problems when you do this: it is very architecture dependent and you might get a bus error. — cup, Feb 20 '14 at 09:30
It's not my code, I just copied this... So, I should avoid it? If so, how to fix the above code to be "proper" c++? — Luka, Feb 20 '14 at 09:32
If you want to know the effect, just run it with some dummy data and see what is generated. If that is what you want, then leave it. If it is not what you want, then change it. Note that out_data is not null terminated. Might be worth adding this after the for loop. — cup, Feb 20 '14 at 09:47

score 1 · Accepted Answer · edited May 23 '17 at 10:31

1

According to the C++03 - no its not legal because it breaks the strict aliasing rule. some compilers like gcc accept this, even without warnings. You will have to convert the chars to shorts using shifting, a class able to handle this instead of the char buffer, or even a union to name a few options.

edited May 23 '17 at 10:31

Community

1
1

answered Feb 20 '14 at 09:22

WeaselFox

7,220
8
44
75

It breaks strict aliasing rules, actually. And array subscript has type `char`, which can be signed. – jrok Feb 20 '14 at 09:24
So, in the end the size of the short array is half the size of the char array? – Luka Feb 20 '14 at 09:26
@jrok - I dont see why those two lines break aliasing rules. To make sure I compiled this with g++. – WeaselFox Feb 20 '14 at 09:31
it messed up ugly when *in_data is negative, which may be sign extended/translated to pwHex[4billion] or pwHex[thousand zillion] – Non-maskable Interrupt Feb 20 '14 at 09:32
@Luka - the size of the short array depends on the sizeof short in your machine. check it – WeaselFox Feb 20 '14 at 09:33
@WeaselFox: Regardless of what GCC does or doesn't say, it is not valid to interpret an object of type `char[]` via a `short`. – Oliver Charlesworth Feb 20 '14 at 09:38
@OliCharlesworth - well, some research shows you are right. I will update the answer. – WeaselFox Feb 20 '14 at 09:42

Is it legal to convert short to char?

1 Answers1