| FazBrowse GitHub Viewer | Trending | | Home |
| Tools: [Download Repo ZIP] [Original HTTPS Page] |
Fixes: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=69350 **`IsADirectoryError`** Fuzzer provided input data can sometimes produce filenames that look like directories and raise `IsADirectoryError` exceptions which crash the fuzzer. This commit catches those cases and returns -1 to instruct libfuzzer that the inputs are not valuable to add to the corpus. **`FileExistsError`** Similar to the above, this is a possible exception case produced by the fuzzed data and not a bug so its handled the same.
Adds a utility function to limit the maximum file name legnth produced by the fuzzer to a max size dictated by the host its run on.
The fuzzer was having trouble analyzing `fuzz_submodule.py` when using the `atheris.instrument_imports()` context manager. Switching to `atheris.instrument_all()` instead slightly increases the startup time for the fuzzer, but significantly improves the fuzzing engines ability to identify new coverage. The changes here also disable warnings that are logged to `stdout` from the SUT. These warnings are expected to happen with some inputs and clutter the fuzzer output logs. They can be optionally re-enabled for debugging by passing a flag o the Python interpreter command line or setting the `PYTHONWARNINGS` environment variable.
There was a problem hiding this comment.
Thanks a lot for the swift fix, it's much appreciated!
Sorry, something went wrong.
Fixes a bug in the `fuzz_submodule` harness where the fuzzed data can produce file names that exceed the maximum size allowed byt the OS. This issue came up previously and was fixed in gitpython-developers#1922, but the submodule file name fixed here was missed in that PR. Fixes: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=69456
| Back | FazBrowse Home | New Git URL |
Fixes: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=69350 (See 7b684cd)
Also: