Update fuzzing coverage steps (#8785)

This commit is contained in:
Andrey Somsikov
2021-11-24 13:13:08 +03:00
committed by GitHub
parent 71b3007a79
commit a2880b81e0

View File

@@ -64,7 +64,7 @@ https://wiki.ith.intel.com/x/2N42bg.
./read_network-fuzzer -max_total_time=600 ./read_network-corpus
```
Consider adding those useful command line options:
- `-jobs=$(nproc)` runs multiple fuzzing jobs in parallel.
- `-jobs=$(nproc)` runs multiple fuzzing jobs in parallel. Note: configuring code coverage profiling with environment variable `LLVM_PROFILE_FILE=deafult-%p.profraw` is required.
- `-rss_limit_mb=0` to ignore out-of-memory issues.
## Analyzing fuzzing quality
@@ -75,7 +75,7 @@ To build coverage report after fuzz test execution run:
```
llvm-profdata merge -sparse *.profraw -o default.profdata && \
llvm-cov show ./read_network-fuzzer -instr-profile=default.profdata -format=html -output-dir=read_network-coverage
llvm-cov show ./read_network-fuzzer -object=lib/libinference_engine.so -instr-profile=default.profdata -format=html -output-dir=read_network-coverage
```
## Reproducing findings