4 ms·
I've been hit by this behaviour just a few days ago with the fantastic libfuzzer, which finds much more errors than afl++ or honggfuzz in much less time. But it
by rurban 6y ago
I've been hit by this behaviour just a few days ago with the fantastic libfuzzer, which finds much more errors than afl++ or honggfuzz in much less time. But it's buffer strategy is enoying, leading to inefficiency and artificial sscanf bugs. Problem is, oss-fuzz/libfuzzer gives you a const char* buffer without zero termination. In reality every such input buffer is zero terminated, because otherwise strtol or sscanf are worthless and lead to buffer overflows. They don't even allow adding a zero at the end, so you have to copy the whole input buffer just to add your zero termination, which makes all the fuzzer speed advantages worthless.