Fletcher-32 checksum
Fletcher-32 checksum. A fast error-detection checksum, not a cryptographic hash.
Runs on your device. The file is never uploaded.
Fletcher-32 checksum reads your file as 16-bit little-endian words, keeps two running sums modulo 65535, and prints eight hex digits below a line naming the tool and the file. An odd-length input borrows an implied zero for its missing high byte. Appending a real zero byte to such a file therefore leaves the digest untouched.
Questions
What does Fletcher-32 read?
Pairs of bytes as 16-bit words, little-endian, with a zero byte assumed if the input has an odd length. The two running sums are taken modulo 65535 and packed into a 32-bit value, printed as eight hex digits.
Why does an odd-length input still work?
The last word is padded with a zero byte. That means a file ending in a zero byte and the same file without it can produce the same checksum, which is a known property of the padding, not a fault in this tool.
How does it compare with CRC32?
Similar width, different trade-off. Fletcher-32 needs no lookup table and is quick in plain arithmetic; CRC32 catches certain burst patterns more reliably and is what zip and gzip use. If you are verifying an archive, you want CRC32.
Is it safe for deduplication?
Only where nobody is trying to trick you. 32 bits means collisions appear by chance across large sets. For content addressing use a wider hash such as BLAKE3 or SHA-256.