diff options
| author | artsdv <[email protected]> | 2026-05-26 23:36:32 +0300 |
|---|---|---|
| committer | artsdv <[email protected]> | 2026-05-27 00:07:51 +0300 |
| commit | 0bf28c535744168a9c29b67aa8108e7b86703ce4 (patch) | |
| tree | 66e53f47ac760fde9470e420430cb0a32d240ac9 /contrib/python/ipython/py3/IPython/utils/importstring.py | |
| parent | e7e308065735cdaf5dcb4348187dd43168e37d8e (diff) | |
change afl logic to afl++
## **AFL\+\+ integration into the fuzzing toolchain (replacing vanilla AFL)**
**Problem being solved:** compile fuzzing targets with the `--afl` flag, substituting the clang/clang\+\+ compiler with afl-clang-fast / afl-clang-fast\+\+.
The current vanilla AFL implementation is unmaintained and broken. The details were discussed here: <https://nda.ya.ru/t/lrUJhJBJ7dSnUG>.
In this PR, the vanilla AFL code has been removed, leaving only `aflpp_driver.cpp` (from the AFLplusplus repository), the code that helps convert libfuzzer-like fuzzing harnesses to afl\+\+. Compilation uses the afl\+\+ toolchain, which was added in advance to `build/external_resources/aflplusplus/`.
I had to make changes to the `devtools/ya` code, since I couldn't find any similar cases (selecting several compilers for a single platform depending on a flag). I chose not to register a new compiler, because afl-clang-fast is essentially a wrapper over clang with added passes — so I decided to select the compiler based on the `--afl` flag.
commit_hash:cc138ebf17f07122cbcfdc9a79b371f6b73cd5ca
Diffstat (limited to 'contrib/python/ipython/py3/IPython/utils/importstring.py')
0 files changed, 0 insertions, 0 deletions