Description
Unclear what's causing it, but upon including filter.c (from the filter pset) in the distribution files and a submission with just filter.c in it too, there's a match with an extremely high score. Weirder still, everything from the distro file was itself ignored as it should and upon inspection 0% of the file matched, but it had a score of 8.8. Which is likely a 100% match, being pulled down a little through common code.
On the run several large files were ignored, and several images too. Removing the distro files did not decrease the number of submissions. Which leads me to believe it's the actual comparison algorithm in the scoring phase that mixes in the distro file with the normal files, and is incorrectly ignoring distro code at that time.