If two known files have the same hash, they are given a match score of 100, regardless of their block size. Here's an example, comparing the output of v2.11.1 with v2.9:
We compute the hash of some file, alpha, and then duplicate it's value, but changing the block size to half its value:
$ cat known.txt
ssdeep,1.1--blocksize:hash:hash,filename
24576:8obBaE9mvOeItYe8xGfkOI9emYzHEsXlmUYiaynYOBKPV/7AO07kSYG0nD4Q:vmOeIKe0G8Ze2sVdqynYOUPVUQ7NsQ,"alpha"
12288:8obBaE9mvOeItYe8xGfkOI9emYzHEsXlmUYiaynYOBKPV/7AO07kSYG0nD4Q:vmOeIKe0G8Ze2sVdqynYOUPVUQ7NsQ,"beta"
Note how the current ssdeep matches them:
$ ssdeep -ax known.txt
known.txt:alpha matches known.txt:beta (100)
known.txt:beta matches known.txt:alpha (100)
But an older version does not:
$ ./ssdeep-2.9 -ax known.txt
known.txt:alpha matches known.txt:beta (0)
known.txt:beta matches known.txt:alpha (0)
This issue was originally reported by Tsukasa, li@livegrid[.]org
A fix for this issue has been checked into SVN.