Correct MD5 sum calculation for non-whole byte bit depths

[ktmf01]

As suggested by Josh Coalson

[JeromeMartinez]

Doesn't it change the computing? I understand that with e.g. 10-bit 11 1111 0001 (decimal negative 15):

• Before (zero padding): 0000 0011 1111 0001
• After (sign extended): 1111 1111 1111 0001

I actually never handled content with non-whole byte bit depths, wondering what is the code in libflac.

[ktmf01]

Yes, it changes the computation. The problem is that what is in the current draft is wrong :)

It does make perfect sense though, because the sign extending keeps the value the same. With zero padding negative numbers become positive, with sign-extending the number remains -15 in your example.

[JeromeMartinez]

The problem is that what is in the current draft is wrong :)

Good catch!