Skip to main content
Мы публикуем частые обновления нашей документации, и перевод этой страницы, возможно, еще выполняется. Актуальные сведения см. в документации на английском языке.

Установка CodeQL CLI в системе непрерывной интеграции

Вы можете настроить систему непрерывной интеграции для выполнения CodeQL CLI, выполнения анализа CodeQL и отправки результатов в GitHub Enterprise Cloud для отображения в виде оповещений code scanning.

Code scanning доступен для всех общедоступных репозиториев в Чтобы использовать code scanning в частном репозитории, принадлежащем организации, необходима лицензия на GitHub Advanced Security. Дополнительные сведения см. в разделе Сведения о GitHub Advanced Security.

About generating code scanning results with CodeQL CLI

Once you've made the CodeQL CLI available to servers in your CI system, and ensured that they can authenticate with GitHub Enterprise Cloud, you're ready to generate data.

You use three different commands to generate results and upload them to GitHub Enterprise Cloud:

  1. database create to create a CodeQL database to represent the hierarchical structure of each supported programming language in the repository.
  2. database analyze to run queries to analyze each CodeQL database and summarize the results in a SARIF file.
  3. github upload-results to upload the resulting SARIF files to GitHub Enterprise Cloud where the results are matched to a branch or pull request and displayed as code scanning alerts.

You can display the command-line help for any command using the --help option.

Note: Uploading SARIF data to display as code scanning results in GitHub Enterprise Cloud is supported for organization-owned repositories with GitHub Advanced Security enabled, and public repositories on For more information, see "Managing security and analysis settings for your repository."

Creating CodeQL databases to analyze

  1. Check out the code that you want to analyze:

    • For a branch, check out the head of the branch that you want to analyze.
    • For a pull request, check out either the head commit of the pull request, or check out a GitHub-generated merge commit of the pull request.
  2. Set up the environment for the codebase, making sure that any dependencies are available. For more information, see "Creating CodeQL databases" and "Creating CodeQL databases."

  3. Find the build command, if any, for the codebase. Typically this is available in a configuration file in the CI system.

  4. Run codeql database create from the checkout root of your repository and build the codebase.

    # Single supported language - create one CodeQL database
    codeql database create <database> --command<build> --language=<language-identifier>
    # Multiple supported languages - create one CodeQL database per language
    codeql database create <database> --command<build> \
          --db-cluster --language=<language-identifier>,<language-identifier>

    Note: If you use a containerized build, you need to run the CodeQL CLI inside the container where your build task takes place.

<database>Specify the name and location of a directory to create for the CodeQL database. The command will fail if you try to overwrite an existing directory. If you also specify --db-cluster, this is the parent directory and a subdirectory is created for each language analyzed.
--languageSpecify the identifier for the language to create a database for, one of: cpp`, `csharp`, `go`, `java`, `javascript`, `python`, and `ruby (use javascript to analyze TypeScript code and java to analyze Kotlin code). When used with --db-cluster, the option accepts a comma-separated list, or can be specified more than once.
--commandRecommended. Use to specify the build command or script that invokes the build process for the codebase. Commands are run from the current folder or, where it is defined, from --source-root. Not needed for Python and JavaScript/TypeScript analysis.
--db-clusterUse in multi-language codebases to generate one database for each language specified by --language.
--no-run-unnecessary-buildsRecommended. Use to suppress the build command for languages where the CodeQL CLI does not need to monitor the build (for example, Python and JavaScript/TypeScript).
--source-rootUse if you run the CLI outside the checkout root of the repository. By default, the database create command assumes that the current directory is the root directory for the source files, use this option to specify a different location.
--codescanning-configAdvanced. Use if you have a configuration file that specifies how to create the CodeQL databases and what queries to run in later steps. For more information, see "Customizing code scanning" and "database create."

For more information, see "Creating CodeQL databases."

Single language example

This example creates a CodeQL database for the repository checked out at /checkouts/example-repo. It uses the JavaScript extractor to create a hierarchical representation of the JavaScript and TypeScript code in the repository. The resulting database is stored in /codeql-dbs/example-repo.

$ codeql database create /codeql-dbs/example-repo --language=javascript \
    --source-root /checkouts/example-repo

> Initializing database at /codeql-dbs/example-repo.
> Running command [/codeql-home/codeql/javascript/tools/autobuild.cmd]
    in /checkouts/example-repo.
> [build-stdout] Single-threaded extraction.
> [build-stdout] Extracting
> Finalizing database at /codeql-dbs/example-repo.
> Successfully created database at /codeql-dbs/example-repo.

Multiple language example

This example creates two CodeQL databases for the repository checked out at /checkouts/example-repo-multi. It uses:

  • --db-cluster to request analysis of more than one language.
  • --language to specify which languages to create databases for.
  • --command to tell the tool the build command for the codebase, here make.
  • --no-run-unnecessary-builds to tell the tool to skip the build command for languages where it is not needed (like Python).

The resulting databases are stored in python and cpp subdirectories of /codeql-dbs/example-repo-multi.

$ codeql database create /codeql-dbs/example-repo-multi \
    --db-cluster --language python,cpp \
    --command make --no-run-unnecessary-builds \
    --source-root /checkouts/example-repo-multi
Initializing databases at /codeql-dbs/example-repo-multi.
Running build command: [make]
[build-stdout] Calling python3 /codeql-bundle/codeql/python/tools/
[build-stdout] Calling python3 -S /codeql-bundle/codeql/python/tools/ -v -z all -c /codeql-dbs/example-repo-multi/python/working/trap_cache -p ERROR: 'pip' not installed.
[build-stdout] /usr/local/lib/python3.6/dist-packages -R /checkouts/example-repo-multi
[build-stdout] [INFO] Python version 3.6.9
[build-stdout] [INFO] Python extractor version 5.16
[build-stdout] [INFO] [2] Extracted file /checkouts/example-repo-multi/ in 5ms
[build-stdout] [INFO] Processed 1 modules in 0.15s
[build-stdout] <output from calling 'make' to build the C/C++ code>
Finalizing databases at /codeql-dbs/example-repo-multi.
Successfully created databases at /codeql-dbs/example-repo-multi.

Analyzing a CodeQL database

  1. Create a CodeQL database (see above).
  2. Run codeql database analyze on the database and specify which packs and/or queries to use.
    codeql database analyze <database> --format=<format> \
        --output=<output>  --download <packs,queries>

Note: If you analyze more than one CodeQL database for a single commit, you must specify a SARIF category for each set of results generated by this command. When you upload the results to GitHub Enterprise Cloud, code scanning uses this category to store the results for each language separately. If you forget to do this, each upload overwrites the previous results.

codeql database analyze <database> --format=<format> \
    --sarif-category=<language-specifier> --output=<output> \
<database>Specify the path for the directory that contains the CodeQL database to analyze.
<packs,queries>Specify CodeQL packs or queries to run. To run the standard queries used for code scanning, omit this parameter. To see the other query suites included in the CodeQL CLI bundle, look in /<extraction-root>/qlpacks/codeql/<language>-queries/codeql-suites. For information about creating your own query suite, see Creating CodeQL query suites in the documentation for the CodeQL CLI.
--formatSpecify the format for the results file generated by the command. For upload to GitHub this should be: sarif-latest. For more information, see "SARIF support for code scanning."
--outputSpecify where to save the SARIF results file.
--sarif-categoryOptional for single database analysis. Required to define the language when you analyze multiple databases for a single commit in a repository.

Specify a category to include in the SARIF results file for this analysis. A category is used to distinguish multiple analyses for the same tool and commit, but performed on different languages or different parts of the code.
--sarif-add-query-helpUse if you want to include any available markdown-rendered query help for custom queries used in your analysis. Any query help for custom queries included in the SARIF output will be displayed in the code scanning UI if the relevant query generates an alert. For more information, see "Analyzing databases with the CodeQL CLI."
<packs>Use if you want to include CodeQL query packs in your analysis. For more information, see "Downloading and using CodeQL packs."
--downloadUse if some of your CodeQL query packs are not yet on disk and need to be downloaded before running queries.
--threadsUse if you want to use more than one thread to run queries. The default value is 1. You can specify more threads to speed up query execution. To set the number of threads to the number of logical processors, specify 0.
--verboseUse to get more detailed information about the analysis process and diagnostic data from the database creation process.

For more information, see Analyzing databases with the CodeQL CLI."

Basic example of analyzing a CodeQL database

This example analyzes a CodeQL database stored at /codeql-dbs/example-repo and saves the results as a SARIF file: /temp/example-repo-js.sarif. It uses --sarif-category to include extra information in the SARIF file that identifies the results as JavaScript. This is essential when you have more than one CodeQL database to analyze for a single commit in a repository.

$ codeql database analyze /codeql-dbs/example-repo  \
    javascript-code-scanning.qls --sarif-category=javascript \
    --format=sarif-latest --output=/temp/example-repo-js.sarif

> Running queries.
> Compiling query plan for /codeql-home/codeql/qlpacks/codeql-javascript/AngularJS/DisablingSce.ql.
> Shutting down query evaluator.
> Interpreting results.

Uploading results to GitHub Enterprise Cloud

You can check that the SARIF properties have the supported size for upload and that the file is compatible with code scanning. For more information, see "SARIF support for code scanning".

Before you can upload results to GitHub Enterprise Cloud, you must determine the best way to pass the GitHub App or personal access token you created earlier to the CodeQL CLI (see Installing CodeQL CLI in your CI system). We recommend that you review your CI system's guidance on the secure use of a secret store. The CodeQL CLI supports:

  • Passing the token to the CLI via standard input using the --github-auth-stdin option (recommended).
  • Saving the secret in the environment variable GITHUB_TOKEN and running the CLI without including the --github-auth-stdin option.

When you have decided on the most secure and reliable method for your CI server, run codeql github upload-results on each SARIF results file and include --github-auth-stdin unless the token is available in the environment variable GITHUB_TOKEN.

echo "$UPLOAD_TOKEN" | codeql github upload-results --repository=<repository-name> \
      --ref=<ref> --commit=<commit> --sarif=<file> \
--repositorySpecify the OWNER/NAME of the repository to upload data to. The owner must be an organization within an enterprise that has a license for GitHub Advanced Security and GitHub Advanced Security must be enabled for the repository, unless the repository is public. For more information, see "Managing security and analysis settings for your repository."
--refSpecify the name of the ref you checked out and analyzed so that the results can be matched to the correct code. For a branch use: refs/heads/BRANCH-NAME, for the head commit of a pull request use refs/pull/NUMBER/head, or for the GitHub-generated merge commit of a pull request use refs/pull/NUMBER/merge.
--commitSpecify the full SHA of the commit you analyzed.
--sarifSpecify the SARIF file to load.
--github-auth-stdinUse to pass the CLI the GitHub App or personal access token created for authentication with GitHub's REST API via standard input. This is not needed if the command has access to a GITHUB_TOKEN environment variable set with this token.

For more information, see "github upload-results."

Basic example of uploading results to GitHub Enterprise Cloud

This example uploads results from the SARIF file temp/example-repo-js.sarif to the repository my-org/example-repo. It tells the code scanning API that the results are for the commit deb275d2d5fe9a522a0b7bd8b6b6a1c939552718 on the main branch.

$ echo $UPLOAD_TOKEN | codeql github upload-results --repository=my-org/example-repo \
    --ref=refs/heads/main --commit=deb275d2d5fe9a522a0b7bd8b6b6a1c939552718 \
    --sarif=/temp/example-repo-js.sarif --github-auth-stdin

There is no output from this command unless the upload was unsuccessful. The command prompt returns when the upload is complete and data processing has begun. On smaller codebases, you should be able to explore the code scanning alerts in GitHub Enterprise Cloud shortly afterward. You can see alerts directly in the pull request or on the Security tab for branches, depending on the code you checked out. For more information, see "Triaging code scanning alerts in pull requests" and "Managing code scanning alerts for your repository."

Downloading and using CodeQL query packs

Note: The CodeQL package management functionality, including CodeQL packs, is currently in beta and subject to change.

The CodeQL CLI bundle includes queries that are maintained by GitHub experts, security researchers, and community contributors. If you want to run queries developed by other organizations, CodeQL query packs provide an efficient and reliable way to download and run queries. For more information, see "About code scanning with CodeQL."

Before you can use a CodeQL pack to analyze a database, you must download any packages you require from the GitHub Container registry. This can be done either by using the --download flag as part of the codeql database analyze command. If a package is not publicly available, you will need to use a GitHub App or personal access token to authenticate. For more information and an example, see "Uploading results to GitHub Enterprise Cloud" above.

<scope/name@version:path>Specify the scope and name of one or more CodeQL query packs to download using a comma-separated list. Optionally, include the version to download and unzip. By default the latest version of this pack is downloaded. Optionally, include a path to a query, directory, or query suite to run. If no path is included, then run the default queries of this pack.
--github-auth-stdinPass the GitHub App or personal access token created for authentication with GitHub's REST API to the CLI via standard input. This is not needed if the command has access to a GITHUB_TOKEN environment variable set with this token.

Note: If you specify a particular version of a query pack to use, be aware that the version you specify may eventually become too old for the latest version of CodeQL to make efficient use of. To ensure optimal performance, if you need to specify exact query pack versions, you should reevaluate which versions you pin to whenever you upgrade the CodeQL CLI you're using.

For more information about pack compatibility, see "Publishing and using CodeQL packs."

Basic example of downloading and using query packs

This example runs the codeql database analyze command with the --download option to:

  1. Download the latest version of the octo-org/security-queries pack.
  2. Download a version of the octo-org/optional-security-queries pack that is compatible with version 1.0.1 (in this case, it is version 1.0.2). For more information on semver compatibility, see npm's semantic version range documentation.
  3. Run all the default queries in octo-org/security-queries.
  4. Run only the query queries/csrf.ql from octo-org/optional-security-queries
$ echo $OCTO-ORG_ACCESS_TOKEN | codeql database analyze --download /codeql-dbs/example-repo \
    octo-org/security-queries \
    octo-org/optional-security-queries@~1.0.1:queries/csrf.ql \
    --format=sarif-latest --output=/temp/example-repo-js.sarif

> Download location: /Users/mona/.codeql/packages
> Installed fresh octo-org/security-queries@1.0.0
> Installed fresh octo-org/optional-security-queries@1.0.2
> Running queries.
> Compiling query plan for /Users/mona/.codeql/packages/octo-org/security-queries/1.0.0/potential-sql-injection.ql.
> [1/2] Found in cache: /Users/mona/.codeql/packages/octo-org/security-queries/1.0.0/potential-sql-injection.ql.
> Starting evaluation of octo-org/security-queries/query1.ql.
> Compiling query plan for /Users/mona/.codeql/packages/octo-org/optional-security-queries/1.0.2/queries/csrf.ql.
> [2/2] Found in cache: /Users/mona/.codeql/packages/octo-org/optional-security-queries/1.0.2/queries/csrf.ql.
> Starting evaluation of octo-org/optional-security-queries/queries/csrf.ql.
> [2/2 eval 694ms] Evaluation done; writing results to octo-org/security-queries/query1.bqrs.
> Shutting down query evaluator.
> Interpreting results.

Direct download of CodeQL packs

If you want to download a CodeQL pack without running it immediately, then you can use the codeql pack download command. This is useful if you want to avoid accessing the internet when running CodeQL queries. When you run the CodeQL analysis, you can specify packs, versions, and paths in the same way as in the previous example:

echo $OCTO-ORG_ACCESS_TOKEN | codeql pack download <scope/name@version:path> <scope/name@version:path> ...

Downloading CodeQL packs from multiple GitHub container registries

If your CodeQL packs reside on multiple container registries, then you must instruct the CodeQL CLI where to find each pack. For more information, see "Customizing code scanning."

Example CI configuration for CodeQL analysis

This is an example of the series of commands that you might use to analyze a codebase with two supported languages and then upload the results to GitHub Enterprise Cloud.

# Create CodeQL databases for Java and Python in the 'codeql-dbs' directory
# Call the normal build script for the codebase: 'myBuildScript'

codeql database create codeql-dbs --source-root=src \
    --db-cluster --language=java,python --command=./myBuildScript

# Analyze the CodeQL database for Java, 'codeql-dbs/java'
# Tag the data as 'java' results and store in: 'java-results.sarif'

codeql database analyze codeql-dbs/java java-code-scanning.qls \
    --format=sarif-latest --sarif-category=java --output=java-results.sarif

# Analyze the CodeQL database for Python, 'codeql-dbs/python'
# Tag the data as 'python' results and store in: 'python-results.sarif'

codeql database analyze codeql-dbs/python python-code-scanning.qls \
    --format=sarif-latest --sarif-category=python --output=python-results.sarif

# Upload the SARIF file with the Java results: 'java-results.sarif'

echo $UPLOAD_TOKEN | codeql github upload-results --repository=my-org/example-repo \
    --ref=refs/heads/main --commit=deb275d2d5fe9a522a0b7bd8b6b6a1c939552718 \
    --sarif=java-results.sarif --github-auth-stdin

# Upload the SARIF file with the Python results: 'python-results.sarif'

echo $UPLOAD_TOKEN | codeql github upload-results --repository=my-org/example-repo \
    --ref=refs/heads/main --commit=deb275d2d5fe9a522a0b7bd8b6b6a1c939552718 \
    --sarif=python-results.sarif --github-auth-stdin

Troubleshooting the CodeQL CLI in your CI system

Viewing log and diagnostic information

When you analyze a CodeQL database using a code scanning query suite, in addition to generating detailed information about alerts, the CLI reports diagnostic data from the database generation step and summary metrics. For repositories with few alerts, you may find this information useful for determining if there are genuinely few problems in the code, or if there were errors generating the CodeQL database. For more detailed output from codeql database analyze, use the --verbose option.

For more information about the type of diagnostic information available, see "Viewing code scanning logs".

Code scanning only shows analysis results from one of the analyzed languages

By default, code scanning expects one SARIF results file per analysis for a repository. Consequently, when you upload a second SARIF results file for a commit, it is treated as a replacement for the original set of data.

If you want to upload more than one set of results to the code scanning API for a commit in a repository, you must identify each set of results as a unique set. For repositories where you create more than one CodeQL database to analyze for each commit, use the --sarif-category option to specify a language or other unique category for each SARIF file that you generate for that repository.

Issues with Python extraction

We are deprecating Python 2 support for the CodeQL CLI, more specifically for the CodeQL database generation phase (code extraction).

If you use the CodeQL CLI to run CodeQL code scanning on code written in Python, you must make sure that your CI system has Python 3 installed.

Further reading