Question for fans of fuzzers and property-based testing: what’s the best way to split the raw fuzzy input into two files, e.g. header file and source file? break at a β€˜\0’ or something more tricky?
3
1
read a byte use it as a ratio of input sizes to split the rest perhaps?
1
Ooh I hadn’t thought of ratios! Does that actually work better than a delimiter in practice?
1
Replying to @fanf
I tend to try for the absolute simplest thing i can in a fuzzer wrapper and the problem with a delimiter is you're saying that the delimiter will therefore never show up in the first input since by definition when you encounter it you switch to the second input.

Jun 1, 2020 Β· 8:28 PM UTC

1
Replying to @dsilverstone
yes this is parsing source code, so I can choose one of several equivalent uninteresting control characters to be the delimiter :-) '\0' is actually a bad choice because it's a special sentinel that is still supposed to work when it appears in the input
1
1
thanks for the helpful ideas :-)
1
1
more replies