corru

A language where every byte is an instruction, so any file is a program, and every program slowly eats itself.

Download corru.c

Byte 00 is the corrupt op: it zeroes the next byte and skips it. A loop that comes back finds the damage.

What it does

Every byte is an op
All 256 byte values do something. A text file, a photo or a damaged download is a valid program, and there is nothing to reject.
Nothing can fail
No syntax errors and no runtime errors. Popping an empty stack gives 0, dividing by zero gives 0, and a run just ends when it reaches the last byte.
Programs corrupt themselves
Op 0 overwrites the next program byte with a null and skips it, so a looping program rewrites its own code as it runs.
Undefined behavior, on purpose
Op 3 makes the next op undefined: it can become another op, truncate the stack, skip ahead or corrupt a random byte. It is seeded with T, so a run can be repeated exactly.
Run any file
Point it at an ELF binary, a .wasm module or a photo and it runs the file’s own bytes as instructions. The same file always gives the same output.
Corrupt a file
-c out saves the program as it looked after its own corruption. It never overwrites the input.
One C file, no flags
No lexer, parser or syntax tree, because the source is already the instruction list. It builds with plain gcc, and the interpreter ran clean under ASAN and UBSAN fuzzing.

Run it here

Loading corru…












Try it


gcc corru.c -o corru
./corru -e 104p101p108p108p111p.        # prints hello
./corru -g 64 > noise.cou; ./corru noise.cou
./corru -c broken.bin some.file         # save the corrupted copy


Flags: -e run a string, -s N step limit, -g N make N random bytes, -c OUT save the corrupted program, -t trim a trailing newline.