Problems with bmp-reading program

I'm having trouble finding the bitmap offset info in the file header, and the size info is a bit unreliable at the moment.

I've made a 10x10 pixel 24-bit image in paint in WinXP and that's the file I'm testing with.

If I increase the size so it goes over the size of an unsigned short int it doesn't show the correct size and I was wondering how to fix this. If I increase the size of bmpFileHeader.size to unsigned long int it doesn't work at all.

The second problem is the bitmap offset, it doesn't show any meaningful value at all.

Code:

#include <iostream>
#include <fstream>

using namespace std;

struct bmpFileHeader {
    unsigned short int type;            //specifies the file type
    unsigned short int size;             //specifies the size in bytes of the bitmap file
    unsigned short int reserved1;       //reserved; must be 0
    unsigned short int reserved2;       //reserved; must be 0
    unsigned long int bitmapOffset;     //specifies the offset in bytes from the bitmapfileheader to the bitmap bits
};

int main(int argc, char *argv[])
{
    bmpFileHeader fh;
    
    //Open file as read only, binary
    ifstream bmpstream("test.bmp", ios::in | ios:: binary);
    
    
    //Verifying that open file succeeded
    if(!bmpstream) {
        cout << "Error opening file.\n";
        system("PAUSE");
        return EXIT_FAILURE;
    }
    
    
    //Read bitmap file header
    bmpstream.read((char *) &fh, sizeof(bmpFileHeader));
    
    
    //Temporary debug feedback
    cout << "Filetype: " << hex << fh.type << endl;
    cout << "Filesize: " << dec << fh.size << endl;
    cout << "Reserved 1: " << fh.reserved1 << endl;
    cout << "Reserved 2: " << fh.reserved2 << endl;
    cout << "Bitmap Offset: " << fh.bitmapOffset << endl;
    
    
    //Verifying that file is .BMP by checking first 2 bytes in the file header.
    if (fh.type !=0x4D42) {
        bmpstream.close();
        return EXIT_FAILURE;
    } 
    
    
    
    //Close file
    bmpstream.close();
    
    system("PAUSE");
    return EXIT_SUCCESS;
}

The sizes go 2, 4, 2, 2, 4. You have 2, 2, 2, 2, 4.

Check this out. http://www.wotsit.org/refer.asp?url=...l&sc=246550348

Which OS/Compiler are you using here?

The further you step away from a 16-bit Microsoft / X86 implementation, the less likely reading directly into a structure is to work as desired.

Binary structures are beset with all sorts of endian and alignment problems.

dwks: Ya, I know. I had it like that first, but it only worked with the size when I changed it to 2, 2, 2, 2, 4.

Salem: I'm using Windows XP and using Dev-C++.

Any ideas on what to do?

btw, I like your avatar Salem

Edit: I forgot my manners, thanks for your answers!

The reason it sort of works with your modified struct is that the compiler will enforce alignment on double-words [such as int] when it builds the struct. This means:
2,4,2,2,4 becomes 2,[2,]4,2,2,4 where the [2,] is a filler padding the next 4 bytes to be aligned to a 4 byte boundary.

You can use some sort of mechanism to change the alignment of the struct, so that it's not padded out [aka "packing the struct"] the data structure - I think MS uses

Code:

#pragma pack(1)

.

Or you could read each little bit at a time.

Unfortunatley, there's really no consistant and portable way to use "structure packing". gcc uses __attribute__(something), if I remember correctly.

--
Mats

dwks: Ya, I know. I had it like that first, but it only worked with the size when I changed it to 2, 2, 2, 2, 4.

In that case, your structure padding might be messing things up. Try changing them to the sizes I suggested and reading in each structure member individually. For example, after changing size to a long, this works for me:

Code:

    bmpstream.read((char *) &fh.type, sizeof(fh.type));
    bmpstream.read((char *) &fh.size, sizeof(fh.size));
    // ...

Of course, that's a pain to write out.

Or mess with your compiler's structure padding settings if you feel like it.

[edit] Also see matsp's post. [/edit]

Woah! It works!

This padding was news to me. I'm not exactly an experienced programmer, just doing it for fun.

I've struggled with it for quite some time, I thought I would figure it out for myself but it seems it's hard to find something you don't know you should look for.

I greatly appreciate your help.

http://www.fastgraph.com/help/bmp_header_format.html

Even "#pragma(pack)" can only do so much. On a different architecture, even the basic types may have different sizes, and then there is the whole endian problem to be solved.

The only portable way is to read one byte at a time, and manually write the code to position that data in the proper place in the struct.