Link Search Menu Expand Document

Code Coverage

For projects written in C/C++, you can generate code coverage reports using Clang source-based code coverage. This page walks you through the basic steps. For more details, see Clang’s documentation.

Code coverage reports generation for other languages is not supported yet.

Pull the latest Docker images

Docker images get regularly updated with a newer version of build tools, build configurations, scripts, and other changes. We recommend you pull the most recent images by running the following command:

$ python infra/ pull_images

Build fuzz targets

Code coverage report generation requires a special build configuration to be used. To create a code coverage build for your project, run these commands:

$ python infra/ build_image $PROJECT_NAME
$ python infra/ build_fuzzers --sanitizer=coverage $PROJECT_NAME

Establish access to GCS

To get a good understanding of fuzz testing quality, you should generate code coverage reports by running fuzz targets against the corpus aggregated by OSS-Fuzz. Set up gsutil and ensure that you have access to the corpora by doing the following:

  • Install the gsutil tool.
  • Check whether you have access to the corpus for your project:
$ gsutil ls gs://${PROJECT_NAME}

If you see an authorization error from the command above, run this:

$ gcloud auth login

and try again. Once gsutil works, you can run the report generation.

Generate code coverage reports

Full project report

If you want to generate a code coverage report using the corpus aggregated on OSS-Fuzz, run this command:

$ python infra/ coverage $PROJECT_NAME

If you want to generate a code coverage report using the corpus you have locally, copy the corpus into the build/corpus/$PROJECT_NAME/<fuzz_target_name>/ directories for each fuzz target, then run this command:

$ python infra/ coverage --no-corpus-download $PROJECT_NAME

Single fuzz target

You can generate a code coverage report for a particular fuzz target by using the --fuzz-target argument:

$ python infra/ coverage --fuzz-target=<fuzz_target_name> $PROJECT_NAME

In this mode, you can specify an arbitrary corpus location for the fuzz target (instead of the corpus downloaded from OSS-Fuzz) by using --corpus-dir:

$ python infra/ coverage --fuzz-target=<fuzz_target_name> \
    --corpus-dir=<my_local_corpus_dir> $PROJECT_NAME

Additional arguments for llvm-cov

You may want to use some of the options provided by the llvm-cov tool, like -ignore-filename-regex=. You can pass these to the helper script after --:

$ python infra/ coverage $PROJECT_NAME -- \
    -ignore-filename-regex=.*code/to/be/ignored/.* <other_extra_args>

If you want to specify particular source files or directories to show in the report, list their paths at the end of the extra arguments sequence:

$ python infra/ coverage zlib -- \
    <other_extra_args> /src/zlib/inftrees.c /src/ /src/zlib/zutil.c

If you want OSS-Fuzz to use extra arguments when generating code coverage reports for your project, add the arguments into your project.yaml file as follows:

coverage_extra_args: -ignore-filename-regex=.*crc.* -ignore-filename-regex=.*adler.* <other_extra_args>