undercover.el

A test coverage library for Emacs Lisp.

Notes

A few important notes about undercover.el:

• it assumes a certain development cycle of your package (using either Cask, Eask or Eldev);
• it doesn't support test coverage for byte-compiled files;
• it is based on edebug and can have some issues with macro coverage;
• it doesn't support Circular Objects.

See the combined usage example and buttercup integration example samples for more information.

Installation

Cask

• Add undercover.el to your Cask file:

  (source gnu)
  (source melpa)
  
  (package-file "awesome-package.el")
  
  (development
    (depends-on "undercover"))

• Before invoking load or require with your package in your test runner (test/test-helper.el / features/support/env.el / etc), call undercover with wildcards that will match your package's source files:

  (when (require 'undercover nil t)
    (undercover "*.el" "awesome-extensions/*.el" (:exclude "awesome-examples.el")))
  
  (require 'awesome-package)

• Add your repository to a coverage reporting service, such as Coveralls or Codecov.

Eldev

• Activate undercover plugin in your Eldev file:

  (eldev-use-plugin 'undercover)

• When running tests on CI server, make sure not to do it in packaged or byte-compiled mode. Or do it twice: once with your project loaded as source code, once as a package. The reason is that undercover.el doesn't work with byte-compiled files.

• Add your repository to a coverage reporting service, such as Coveralls or Codecov.

See relevant documentation on Eldev's own page for more information.

Eask

Eask is very similar to Cask; anything that applies to Cask will apply to Eask

• Add undercover.el to your Eask file:

  (package-file "awesome-package.el")
  
  (source "gnu")
  (source "melpa")
  
  (development
    (depends-on "undercover"))

• Before invoking load or require with your package in your test runner (test/test-helper.el / features/support/env.el / etc), call undercover with wildcards that will match your package's source files:

  (when (require 'undercover nil t)
    (undercover "*.el" "awesome-extensions/*.el" (:exclude "awesome-examples.el")))
  
  (require 'awesome-package)

• Add your repository to a coverage reporting service, such as Coveralls or Codecov.

See relevant documentation on Eask's own page for more information.

Configuration

Online services

undercover supports a wide range of combinations of code forges / CI providers / report formats / upload methods. Some common combinations are documented below:

	GitHub Actions	Travis CI	(other CI)
Coveralls	• With GitHub action • With built-in uploader	• With built-in uploader	• With built-in uploader
Codecov	• With GitHub action • With bash uploader	• With bash uploader
(other coverage service)	• Manual configuration		• Manual configuration

---

GitHub Actions + Coveralls + Coveralls GitHub Action

Steps:

1. Add the Coveralls GitHub action to your GitHub Actions workflow YAML file, after your test invocation.

   To support matrix builds, add a final job with parallel-finished: true, as described in the action's documentation.

2. Invoke undercover with (:report-format 'lcov) (:send-report nil).

A complete minimal example (using ert + Cask + ert-runner) can be found here.

---

GitHub Actions + Coveralls + undercover.el built-in uploader

You will need to export the GitHub Actions access token into the environment.

To enable Coveralls parallel builds, set COVERALLS_PARALLEL in the shell environment, and add a final job with if: always() which pings the webhook.

Here is a complete example:

on: [ push, pull_request ]
jobs:
  test:
    runs-on: ubuntu-latest
    strategy:
      matrix:
        emacs_version:
        - 25.3
        - 26.3
        - 27.1
        - snapshot
    steps:
    - uses: purcell/setup-emacs@master
      with:
        version: ${{ matrix.emacs_version }}
    - uses: conao3/setup-cask@master
    - uses: actions/checkout@v2
    - name: Test
      env:
        COVERALLS_FLAG_NAME: Emacs ${{ matrix.emacs_version }}
        COVERALLS_PARALLEL: 1
        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
      run: |
        cask install
        cask exec ert-runner
  finalize:
    runs-on: ubuntu-latest
    if: always()
    needs: test
    steps:
    - run: curl "https://coveralls.io/webhook?repo_name=$GITHUB_REPOSITORY&repo_token=${{ secrets.GITHUB_TOKEN }}" -d "payload[build_num]=$GITHUB_RUN_NUMBER&payload[status]=done"

Alternatively to exporting GITHUB_TOKEN, you may instead specify COVERALLS_REPO_TOKEN, as with any other CI service.

---

Travis CI + Coveralls

No configuration necessary.

To enable Coveralls parallel builds, set COVERALLS_PARALLEL in the shell environment, and configure the web hook as described in the Coveralls documentation.

---

(other CI)

undercover.el has basic support (for reading and parsing relevant environment variables, such as build ID) for the following CI services:

• GitHub Actions
• Travis CI
• Shippable
• Drone
• Jenkins
• Circle CI
• CodeShip
• Wercker
• GitLab CI
• AppVeyor
• Surf
• BuildKite
• Semaphore
• Codefresh

Detected values may be overridden by setting the following environment variables:

• UNDERCOVER_CI_TYPE
• UNDERCOVER_CI_NAME
• UNDERCOVER_COMMIT
• UNDERCOVER_REF
• UNDERCOVER_PULL_REQUEST
• UNDERCOVER_BUILD_ID
• UNDERCOVER_BUILD_NUMBER
• UNDERCOVER_JOB_ID
• UNDERCOVER_JOB_NUMBER
• UNDERCOVER_JOB_NAME

See the documentation of undercover--detect-ci for a description of the semantics for these variables.

---

(other CI) + Coveralls

For CI services which are not "magically" supported by Coveralls, you will need to set the COVERALLS_REPO_TOKEN environment variable before running tests, for example:

$ COVERALLS_REPO_TOKEN=<your-coveralls-repo-token> cask exec ert-runner

Consult the Coveralls documentation for details.

The token should not be made public, so it should be placed in the CI service's secrets store.

Fields in the submitted Coveralls report may be overridden using standard environment variables:

• COVERALLS_SERVICE_NAME
• COVERALLS_REPO_TOKEN
• COVERALLS_SERVICE_NUMBER
• COVERALLS_SERVICE_JOB_ID
• COVERALLS_SERVICE_PULL_REQUEST
• COVERALLS_PARALLEL
• COVERALLS_FLAG_NAME
• COVERALLS_RUN_AT

See the Coveralls API reference for a description of these fields.

---

GitHub Actions + Codecov + CodeCov GitHub Action

Steps:

1. Enable the Codecov app for your account / organization / repository.

2. Add the Codecov GitHub action to your GitHub Actions workflow YAML file, after your test invocation.

3. Invoke undercover with (:report-format 'codecov) (:send-report nil).

A complete minimal example (using ert + Cask + ert-runner) can be found here.

---

(other CI) + Codecov

Codecov is supported in combination with their bash upload script.

In your test runner:

(undercover "*.el" (:report-format 'codecov)
                   (:send-report nil))

And in your pipeline (.travis.yml or equivalent):

after_success:
  # Upload coverage
  - bash <(curl -s https://codecov.io/bash)

(other coverage service)

If the coverage service supports coverage reports in the LCOV, SimpleCov, or Coveralls file format, it should be usable with undercover as follows:

1. When calling undercover, set :report-format to an appropriate compatible format
2. Set :send-report nil
3. If necessary, set :report-file to the location where the report file should be saved
4. Consult the coverage service's documentation on how to upload the report file to their service.

Local reports

Cask / Eask / Emacs Lisp

• Set the report-file option to change the report file location:

  (undercover "*.el" (:report-file "/tmp/local-report.json"))

  undercover.el will try to merge new report with existing one.

• Set the send-report option to nil to disable uploading the coverage report to an online service:

  (undercover "*.el" (:report-file "/tmp/local-report.json")
                     (:send-report nil))

• Set report-format to use a different format for the report file:

  (undercover "*.el" (:report-file "coverage/.resultset.json")
                     (:report-format 'simplecov)
                     (:send-report nil))

  See the documentation of the undercover function for more details.

• Set the UNDERCOVER_FORCE environment variable to calculate coverage even when not running on a CI:

  $ UNDERCOVER_FORCE=true cask exec ert-runner

  Alternatively, use (setq undercover-force-coverage t) before calling (undercover).

• Set the UNDERCOVER_CONFIG variable to configure undercover.el via the environment:

  (when (require 'undercover nil t) (undercover))

  $ UNDERCOVER_CONFIG='("*.el" (:exclude "awesome-examples.el"))' cask exec ert-runner

Eldev

With Eldev generating local reports is very easy:

$ eldev test -U local-report.json

Option -U is the short form of --undercover-report and is only available if the plugin is activated (see above). Option -u (--undercover) lets you configure the library from command line. For more information see Eldev's documentation.

Selecting which exactly files you want undercover.el to instrument is not possible from command line: these always default to all .el files in main fileset. However, you can adjust variable eldev-undercover-fileset in file Eldev if you need to change that for some reason.

Viewing coverage in Emacs

Simple report

You can generate a simple coverage summary report using the 'text report format:

(require 'undercover)
(setq undercover-force-coverage t)
(undercover "*.el" (:report-file nil) ; or a text file to save the report to
                   (:report-format 'text))

Coverage overlay

1. Install coverage-mode

2. Create the coverage directory in your project root

3. Configure undercover.el as follows:

   (require 'undercover)
   (setq undercover-force-coverage t)
   (undercover "*.el" (:report-format 'simplecov)
                      (:send-report nil))

4. Run your tests

5. Open a source file, and enable coverage-mode.

Troubleshooting

Code in macros is not included in coverage reports

You may need to teach edebug how to instrument the affected macros.

See "Instrumenting Macro Calls" in the Emacs manual and the documentation of def-edebug-spec for more details.

"UNDERCOVER: Error while loading ... for coverage:"

The problem may be due to edebug failing to parse the file. Follow the instructions under the error message.

"UNDERCOVER: No coverage information [...]"

Try the following:

1. remove byte-compiled files (*.elc) of your project
2. load and configure undercover before your project files (see above)
3. make sure ert-runner does not load your project files (your project's .ert-runner should use -L instead of -l for files you want to measure coverage of)