Do those exercise the logic kazinator called out as test-worthy? To my eye, no. They don't use search_file, and their inputs are smaller than MIN_FILE_SIZE_FOR_THREADS anyway.
I'm inclined to agree with kazinator. The code here: <https://github.com/davidesantangelo/krep/blob/ac6783af42c92f...> looks wrong to me. It potentially increases `chunk_size` but doesn't reduce the number of loop iterations to be consistent with that. Maybe search_thread recognizes the boundary violation and does nothing? That'd be a relatively harmless outcome, but it's strange to launch threads that do nothing. But actually it's not immediately obvious to me that it does recognize if end_pos is beyond file_len. And then the code about handling + skipping overlaps in search_thread also looks test-worthy.
