grcov

grcov collects and aggregates code coverage information for multiple source files. grcov processes .profraw and .gcda files which can be generated from llvm/clang or gcc. grcov also processes lcov files (for JS coverage) and JaCoCo files (for Java coverage). Linux, macOS and Windows are supported.

This is a project initiated by Mozilla to gather code coverage results on Firefox.

Table of Contents

• man grcov
• How to get grcov
• Usage
  • Example: How to generate source-based coverage for a Rust project
  • Example: How to generate .gcda files for C/C++
  • Example: How to generate .gcda files for a Rust project
  • Generate a coverage report from coverage artifacts
    • LCOV output
    • Coveralls output
    • grcov with Travis
    • grcov with Gitlab
  • Alternative reports
  • Hosting HTML reports and using coverage badges
    • Example
  • Enabling symlinks on Windows
• Auto-formatting
• Build & Test
• Minimum requirements
• License

man grcov

Usage: grcov [OPTIONS] <PATHS>...

Arguments:
  <PATHS>...
          Sets the input paths to use

Options:
  -b, --binary-path <PATH>
          Sets the path to the compiled binary to be used

      --llvm-path <PATH>
          Sets the path to the LLVM bin directory

  -t, --output-types <OUTPUT TYPE>
          Comma separated list of custom output types:
          - *html* for a HTML coverage report;
          - *coveralls* for the Coveralls specific format;
          - *lcov* for the lcov INFO format;
          - *covdir* for the covdir recursive JSON format;
          - *coveralls+* for the Coveralls specific format with function information;
          - *ade* for the ActiveData-ETL specific format;
          - *files* to only return a list of files.
          - *markdown* for human easy read.
          - *cobertura* for output in cobertura format.


          [default: lcov]

  -o, --output-path <PATH>
          Specifies the output path. This is a file for a single output type and must be a folder
          for multiple output types

      --output-config-file <PATH>
          Specifies the output config file

  -s, --source-dir <DIRECTORY>
          Specifies the root directory of the source files

  -p, --prefix-dir <PATH>
          Specifies a prefix to remove from the paths (e.g. if grcov is run on a different machine
          than the one that generated the code coverage information)

      --ignore-not-existing
          Ignore source files that can't be found on the disk

      --ignore <PATH>
          Ignore files/directories specified as globs

      --keep-only <PATH>
          Keep only files/directories specified as globs

      --path-mapping <PATH>


      --branch
          Enables parsing branch coverage information

      --filter <FILTER>
          Filters out covered/uncovered files. Use 'covered' to only return covered files,
          'uncovered' to only return uncovered files

          [possible values: covered, uncovered]

      --llvm
          Speeds-up parsing, when the code coverage information is exclusively coming from a llvm
          build

      --token <TOKEN>
          Sets the repository token from Coveralls, required for the 'coveralls' and 'coveralls+'
          formats

      --commit-sha <COMMIT HASH>
          Sets the hash of the commit used to generate the code coverage data

      --service-name <SERVICE NAME>
          Sets the service name

      --service-number <SERVICE NUMBER>
          Sets the service number

      --service-job-id <SERVICE JOB ID>
          Sets the service job id

          [aliases: service-job-number]

      --service-pull-request <SERVICE PULL REQUEST>
          Sets the service pull request number

      --parallel
          Sets the build type to be parallel for 'coveralls' and 'coveralls+' formats

      --threads <NUMBER>


      --precision <NUMBER>
          Sets coverage decimal point precision on output reports

          [default: 2]

      --guess-directory-when-missing


      --vcs-branch <VCS BRANCH>
          Set the branch for coveralls report. Defaults to 'master'

          [default: master]

      --log <LOG>
          Set the file where to log (or stderr or stdout). Defaults to 'stderr'

          [default: stderr]

      --log-level <LEVEL>
          Set the log level

          [default: ERROR]
          [possible values: OFF, ERROR, WARN, INFO, DEBUG, TRACE]

      --excl-line <regex>
          Lines in covered files containing this marker will be excluded

      --excl-start <regex>
          Marks the beginning of an excluded section. The current line is part of this section

      --excl-stop <regex>
          Marks the end of an excluded section. The current line is part of this section

      --excl-br-line <regex>
          Lines in covered files containing this marker will be excluded from branch coverage

      --excl-br-start <regex>
          Marks the beginning of a section excluded from branch coverage. The current line is part
          of this section

      --excl-br-stop <regex>
          Marks the end of a section excluded from branch coverage. The current line is part of this
          section

      --no-demangle
          No symbol demangling

  -h, --help
          Print help (see a summary with '-h')

  -V, --version
          Print version

How to get grcov

Grcov can be downloaded from releases or, if you have Rust installed, you can run cargo install grcov.

Usage

Example: How to generate source-based coverage for a Rust project

1. Install the llvm-tools or llvm-tools-preview component:

   rustup component add llvm-tools-preview
   

2. Ensure that the following environment variable is set up:

   export RUSTFLAGS="-Cinstrument-coverage"
   

3. Build your code:

   cargo build

4. Ensure each test runs gets its own profile information by defining the LLVM_PROFILE_FILE environment variable (%p will be replaced by the process ID, and %m by the binary signature):

   export LLVM_PROFILE_FILE="your_name-%p-%m.profraw"
   

5. Run your tests:

   cargo test

In the CWD, you will see a .profraw file has been generated. This contains the profiling information that grcov will parse, alongside with your binaries.

Example: How to generate .gcda files for C/C++

Pass --coverage to clang or gcc (or for older gcc versions pass -ftest-coverage and -fprofile-arcs options (see gcc docs).

Example: How to generate .gcda files for a Rust project

Nightly Rust is required to use grcov for Rust gcov-based coverage. Alternatively, you can export RUSTC_BOOTSTRAP=1, which basically turns your stable rustc into a Nightly one.

1. Ensure that the following environment variables are set up:

   export CARGO_INCREMENTAL=0
   export RUSTFLAGS="-Zprofile -Ccodegen-units=1 -Copt-level=0 -Clink-dead-code -Coverflow-checks=off -Zpanic_abort_tests -Cpanic=abort"
   export RUSTDOCFLAGS="-Cpanic=abort"
   

   These will ensure that things like dead code elimination do not skew the coverage.

2. Build your code:

   cargo build

   If you look in target/debug/deps dir you will see .gcno files have appeared. These are the locations that could be covered.

3. Run your tests:

   cargo test

   In the target/debug/deps/ dir you will now also see .gcda files. These contain the hit counts on which of those locations have been reached. Both sets of files are used as inputs to grcov.

Generate a coverage report from coverage artifacts

Generate a html coverage report like this:

grcov . -s . --binary-path ./target/debug/ -t html --branch --ignore-not-existing -o ./target/debug/coverage/

N.B.: The --binary-path argument is only necessary for source-based coverage.

You can see the report in target/debug/coverage/index.html.

(or alternatively with -t lcov grcov will output a lcov compatible coverage report that you could then feed into lcov's genhtml command).

LCOV output

By passing -t lcov you could generate an lcov.info file and pass it to genhtml:

genhtml -o ./target/debug/coverage/ --show-details --highlight --ignore-errors source --legend ./target/debug/lcov.info

LCOV output should be used when uploading to Codecov, with the --branch argument for branch coverage support.

Coveralls output

Coverage can also be generated in coveralls format:

grcov . --binary-path ./target/debug/ -t coveralls -s . --token YOUR_COVERALLS_TOKEN > coveralls.json

grcov with Travis

Here is an example of .travis.yml file for source-based coverage:

language: rust

before_install:
  - curl -L https://github.com/mozilla/grcov/releases/latest/download/grcov-x86_64-unknown-linux-gnu.tar.bz2 | tar jxf -

matrix:
  include:
    - os: linux
      rust: stable

script:
    - rustup component add llvm-tools-preview
    - export RUSTFLAGS="-Cinstrument-coverage"
    - cargo build --verbose
    - LLVM_PROFILE_FILE="your_name-%p-%m.profraw" cargo test --verbose
    - ./grcov . --binary-path ./target/debug/ -s . -t lcov --branch --ignore-not-existing --ignore "/*" -o lcov.info
    - bash <(curl -s https://codecov.io/bash) -f lcov.info

Here is an example of .travis.yml file:

language: rust

before_install:
  - curl -L https://github.com/mozilla/grcov/releases/latest/download/grcov-x86_64-unknown-linux-gnu.tar.bz2 | tar jxf -

matrix:
  include:
    - os: linux
      rust: stable

script:
    - export CARGO_INCREMENTAL=0
    - export RUSTFLAGS="-Zprofile -Ccodegen-units=1 -Copt-level=0 -Clink-dead-code -Coverflow-checks=off -Zpanic_abort_tests -Cpanic=abort"
    - export RUSTDOCFLAGS="-Cpanic=abort"
    - cargo build --verbose $CARGO_OPTIONS
    - cargo test --verbose $CARGO_OPTIONS
    - |
      zip -0 ccov.zip `find . \( -name "YOUR_PROJECT_NAME*.gc*" \) -print`;
      ./grcov ccov.zip -s . -t lcov --llvm --branch --ignore-not-existing --ignore "/*" -o lcov.info;
      bash <(curl -s https://codecov.io/bash) -f lcov.info;

grcov with Gitlab

Here is an example .gitlab-ci.yml which will build your project, then collect coverage data in a format that Gitlab understands. It is assumed that you'll use an image which already has relevant tools installed, if that's not the case put the appropriate commands at the beginning of the script stanza.

build:
  variables:
    # Set an environment variable which causes LLVM to write coverage data to the specified location. This is arbitrary, but the path passed to grcov (the first argument) must contain these files or the coverage data won't be noticed.
    LLVM_PROFILE_FILE: "target/coverage/%p-%m.profraw"
  script:
    # Run all your Rust-based tests
    - cargo test --workspace
    # Optionally, run some other command that exercises your code to get more coverage:
    - ./bin/integration-tests --foo bar
    # Create the output directory
    - mkdir target/coverage
    # This is a multi-line command. You can also write it all as one line if desired, just remove
    # the '|' and all the newlines.
    - |
        grcov
    # This path must match the setting in LLVM_PROFILE_FILE. If you're not getting the coverage
    # you expect, look for '.profraw' files in other directories.
        target/coverage
    # If your target dir is modified, this will need to match...
        --binary-path target/debug
    # Where the source directory is expected
        -s .
    # Where to write the output; this should be a directory that exists.
        -o target/coverage
    # Exclude coverage of crates and Rust stdlib code. If you get unexpected coverage results from
    # this (empty, for example), try different combinations of '--ignore-not-existing',
    # '--ignore "$HOME/.cargo/**"' and see what kind of filtering gets you the coverage you're
    # looking for.
        --keep-only 'src/*'
    # Doing both isn't strictly necessary, if you won't use the HTML version you can modify this
    # line.
        --output-types html,cobertura

    # Extract just the top-level coverage number from the XML report.
    - xmllint --xpath "concat('Coverage: ', 100 * string(//coverage/@line-rate), '%')" target/coverage/cobertura.xml
  coverage: '/Coverage: \d+(?:\.\d+)?/'
  artifacts:
    paths:
      - target/coverage/
    reports:
      coverage_report:
        coverage_format: cobertura
        path: target/coverage.xml

This also ties into Gitlab's coverage percentage collection, so in merge requests you'll be able to see:

• increases or decreases of coverage
• whether particular lines of code modified by a merge request are covered or not.

Additionally, the HTML-formatted coverage report (if you leave it enabled) will be produced as an artifact.

Alternative reports

grcov provides the following output types:

Hosting HTML reports and using coverage badges

The HTML report can be hosted on static website providers like GitHub Pages, Netlify and others. It is common to provide a coverage badge in a project's readme to show the current percentage of covered code.

To still allow adding the badge when using a static site host, grcov generates coverage badges and a JSON file with coverage information that can be used with https://shields.io to dynamically generate badges.

The coverage data for https://shields.io can be found at /coverage.json and the generated bagdes are available as SVGs at /badges/*svg.

The design of generated badges is taken from shields.io but may not be updated immediately if there is any change. Using their endpoint method is recommended if other badges from their service are used already.

Enabling symlinks on Windows

grcov uses symbolic links to avoid copying files, when processing directories of coverage data. On Windows, by default, creating symbolic links to files requires Administrator privileges. (The reason is to avoid security attacks in applications that were designed before Windows added support for symbolic links.)

When running on Windows grcov will attempt to create a symbolic link. If that fails then grcov will fall back to copying the file. Copying is less efficient but at least allows users to run grcov. grcov will also print a warning when it falls back to copying a file, advising the user either to enable the privilege for their account or to run as Administrator.

You can enable the "Create Symbolic Links" privilege for your account so that you do not need to run as Administrator to use grcov.

1. Click Start, then select "Local Group Policy Editor". Or just run gpedit.msc to open it directly.
2. In the navigation tree, select "Computer Configuration", "Windows Settings", "Security Settings", "Local Policies".
3. In the pane on the right, select "Create symbolic links" and double-click it.
4. Click "Add User or Group", and add your account.
5. Log out and then log back in.

Example

Let's consider we have a project at with username sample and project awesome that is hosted with GitHub Pages at https://sample.github.io/awesome.

By using the the shields.io endpoint we can create a Markdown badge like so:

[![coverage](https://shields.io/endpoint?url=https://sample.github.io/awesome/coverage.json)](https://sample.github.io/awesome/index.html)

If we want to avoid using shields.io as well, we can use the generated badges as follows (note the different URL for the image):

[![coverage](https://sample.github.io/awesome/badges/flat.svg)](https://sample.github.io/awesome/index.html)

Auto-formatting

This project is using pre-commit. Please run pre-commit install to install the git pre-commit hooks on your clone. Instructions on how to install pre-commit can be found here.

Every time you will try to commit, pre-commit will run checks on your files to make sure they follow our style standards and they aren't affected by some simple issues. If the checks fail, pre-commit won't let you commit.

Build & Test

Build with:

cargo build

To run unit tests:

cargo test --lib

To run integration tests, it is suggested to use the Docker image defined in tests/Dockerfile. Simply build the image to run them:

docker build -t marcocas/grcov -f tests/Dockerfile .

Otherwise, if you don't want to use Docker, the only prerequisite is to install GCC 7, setting the GCC_CXX environment variable to g++-7 and the GCOV environment variable to gcov-7. Then run the tests with:

cargo test

Minimum requirements

• GCC 4.9 or higher is required (if parsing coverage artifacts generated by GCC).
• Rust 1.52

License

Published under the MPL 2.0 license.