Name selection must be deterministic in order to actually test the tool

[ande2101]

The only way to test whether a book has been correctly deloused is if two different inputs produce the same output. Having random names in there removes the ability for a user to perform this check, thus making the script vulnerable to more subtle watermarking schemes in the future.

So the random module has to go.

[grayleonard]

Yes, you are right. Commit e630fff adds deterministic file naming, which I have just pushed to master.

I found that built zip files have different checksums because of header differences (modification date, filename, etc) - I have included diff.sh to compare two zip files by extracting their contents and diffing each extracted file.