Add fuzzing harnesses for classes CBlockHeader, CFeeRate and various functions.
To test this PR:
$ make distclean
$ ./autogen.sh
$ CC=clang CXX=clang++ ./configure --enable-fuzz \
--with-sanitizers=address,fuzzer,undefined
$ make
$ src/test/fuzz/block_header
^c (ctrl-c)
$ src/test/fuzz/fee_rate
^c (ctrl-c)
$ src/test/fuzz/integer
^c (ctrl-c)
$ src/test/fuzz/multiplication_overflow
^c (ctrl-c)
$ src/test/fuzz/string
^c (ctrl-c)
<!--e57a25ab6845829454e8d69fc972939a-->
The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.
<!--174a7506f384e20aa4161008e828411d-->
Reviewers, this pull request conflicts with the following ones:
If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first.
16 | +{ 17 | + const T i = fuzzed_data_provider.ConsumeIntegral<T>(); 18 | + const T j = fuzzed_data_provider.ConsumeIntegral<T>(); 19 | + const bool is_multiplication_overflow_custom = MultiplicationOverflow(i, j); 20 | + T result_builtin; 21 | + const bool is_multiplication_overflow_builtin = __builtin_mul_overflow(i, j, &result_builtin);
does this compile for afl?
Yes it should :) Let me know if you find any evidence to the contrary :)
ACK 44abf417eb1cd8598084eee1a429ca57c7d0579a 🏉
<details><summary>Show signature and timestamp</summary>
Signature:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512
ACK 44abf417eb1cd8598084eee1a429ca57c7d0579a 🏉
-----BEGIN PGP SIGNATURE-----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=b0BC
-----END PGP SIGNATURE-----
Timestamp of file with hash 0aac1971bcd3d59c21efc537215377f811eface51e1bcab3cac5dcd6d853c5b4 -
</details>
88 | + return i != 0 && (j < (std::numeric_limits<T>::max() / i)); 89 | + } 90 | + } 91 | + } else { 92 | + return j != 0 && i > std::numeric_limits<T>::max() / j; 93 | + }
Why implement this from scratch? Could just use __builtin_mul_overflow?
I think it would be handy to have MultiplicationOverflow(…) in other non-test parts of the code base.
When we are confident of the equivalence between MultiplicationOverflow and __builtin_mul_overflow we can implement MultiplicationOverflow by using the __builtin_mul_overflow builtin where available (gcc + clang) and the current implementation where it is not (MSVC).
That's why I added a fuzz test for MultiplicationOverflow too.
And when we have a version of MultiplicationOverflow(…) that is known to be working on all platforms we can start using it in non-test code.
Makes sense? :)
When talking about integer overflows:
Don't miss this one: a signed integer overflow in Microsoft's SafeInt ("SafeInt is a class library for C++ that manages integer overflows")...
It's like rain on your wedding day It's a free ride when you've already paid It's the good advice that you just didn't take
:)