How to compute the Adler32 in zlib

Question

Regarding the checksum, adler32, used in zlib.

From rfc 1950:

ADLER32: This contains a checksum value of the uncompressed data (excluding any dictionary data) computed according to Adler-32 algorithm.

Here is a png (zlib in IDAT).

• blue = length, 4 bytes
• red = type, 4 bytes
• yellow = crc, 4 bytes
• green = zlib header, 2 bytes
• gray = zlib data, varying bytes
• pink = adler32, 4 bytes

Now parsing the gray data out and passing the zlib data to an external program that calculates the adler32.

The result is 64c60bbd, not b0337596.

---

...or when passing the data from the image map.

The result is 1a49bac4, not b0337596

---

The code of checksum.exe:

const uint32_t MOD_ADLER = 65521;  
uint32_t adler32(unsigned char *data, size_t len) {
  uint32_t a = 1, b = 0;
  size_t index;  
  for (index = 0; index < len; ++index) {
    a = (a + data[index]) % MOD_ADLER;
    b = (b + a) % MOD_ADLER;
  }
  return (b << 16) | a;
}

int main(int argc, const char* argv[]) {
  if(argc == 2) {
    ifstream ifs(argv[1]);
    char c;
    unsigned char data[50000];
    int len = 0;

    while( ifs.get(c) )
      data[len++] = c;

    if( len >= 50000 )
      cout<<"Error: array overflow.\n";

    uint32_t result = adler32(data, len);
    cout<< "\n" << std::hex << result << "\n" ;
    return 0;
  }
}

Question, what could cause a discrepancy in calculation(s)? i.e. what exactly is the input data for zlib's adler32?

Solution:

As stated, the zlib adler32 is calculated by the uncompressed data stream, which in a png is the data stream after filtering. In this case that would be...

Note, the uncompressed data stream here is a red 50x50 image as rgba with an addition byte per row for the filter, resulting in 10,050 bytes. That is 50*50*40 + 50.