Unofficial scripts and Docker configuration for building CodeQL databases for the OpenJDK.
The created databases can then for example be loaded and analyzed using the Visual Studio Code CodeQL extension.
- OS: Windows 10, Linux (not tested)
- CPU architecture: 64-bit
- Docker (Docker Desktop)
- RAM: 8GB or more
See also OpenJDK Build Hardware Requirements.
This project provides convenience scripts for creating a CodeQL database for the Java code of the OpenJDK:
- Windows:
build_database.cmd
- Linux:
build_database.sh
At the moment they use CodeQL CLI 2.15.5 and build a Java database for the latest https://github.com/openjdk/jdk21u commit.
The scripts only build the Linux variant of the JDK, building the Windows variant is currently not supported.
These scripts can be executed as is (assuming that Docker has already been started). They perform the following tasks:
- Build the Docker image (named
codeql-jdk
) - Clone the JDK source code
- Build the CodeQL database and copy it to the
databases
folder of the current directory
The jdk-versions
folder contains scripts for building databases for specific JDK versions.
Note: Building the Docker image, the JDK and the CodeQL database are all resource and time intensive tasks. In total they might take up to an hour (depends on your network connection and hardware).
ℹ️ 3 to 4GB of memory might suffice for the Docker container, however a memory limit should be specified for the JDK build using --memory-limit
(see "Build configuration" section), otherwise the build can get stuck and fail.
The Dockerfile uses build-time variables for configuration.
BOOT_JDK_URL
: URL of the JDK to be used as boot JDK. The URL should be a download link for a.tar.gz
file containing all JDK files nested in an extra directory. For example the Eclipse Adoptium download URLs of the GitHub releases can be used.CODEQL_CLI_VERSION
: Version of CodeQL CLI to use for building the database, e.g.2.5.7
The Docker image has a build script as entry point which allows customizing how the JDK and the CodeQL database is built.
The arguments are passed as additional arguments to docker container run
.
Additionally the arguments can be used with the convenience scripts mentioned in the "Usage" section.
Arguments have the format --param value
--jdk-git-repo
(required)
URI of the Git repository from which the JDK source code should be cloned. When choosing the JDK version to build, the following has to be considered:- A matching boot JDK has to be choosen (see "Docker image configuration" section)
- CodeQL CLI has to support the Java version, see the CodeQL documentation for supported versions. The CodeQL CLI version might have to be adjusted (see "Docker image configuration" section).
CodeQL CLI might not support building the latest (unreleased) JDK version yet, prefer JDK update releases of older versions such as jdk16u. - The Dockerfile is currently configured for the default JDK built by the convenience scripts of this project. Other JDKs might have different dependencies, consult the JDK build instructions and adjust the Dockerfile if problems occur.
- Since the JDK build tools are part of the JDK repository, the choice of the JDK version affects which of the other build arguments are supported and how they behave.
--jdk-commit-sha
Git commit hash (or branch name) of the JDK commit to build. If not specified, the latest commit of the active branch of the remote Git repository is built.
See also the considerations for picking the JDK version described above for the--jdk-git-repo
parameter. When using a commit hash, the full commit hash (40 characters) should be specified to allow performing a shallow fetch.--memory-limit
Specifies the memory limit in MB for the JDK build within the container. The JDK build tools will use the maximum memory available to the container if not specified. It is recommended to specify a custom limit because the JDK build tools do not account for CodeQL CLI running during the build, causing the JDK build to slow down or even fail. Creating a container with ~4GB, and setting a memory limit of ~2GB for the JDK build seems to work fine.
Also have a look at the JDK Build Hardware Requirements and Build Performance guide. When using WSL2 on Windows, tuning the WSL 2 Settings might help as well.--cpu-cores
Specifies the number of CPU cores the JDK build is allowed to use. The JDK build tools will use all cores dedicated to the container by default. Also have a look at the JDK Build Performance guide. When using WSL2 on Windows, tuning the WSL 2 Settings might help as well.--make-target
Specifies themake
target for the JDK build. By default thejava
target is executed, compiling all Java code of the JDK.--codeql-db-lang
Specifies the programming language for which CodeQL CLI shoud create the database, have a look at the CodeQL CLI documentation for a list of supported programming languages. By default the database is created for Java source code. When choosing a different programming language it might be necessary to specify a different--make-target
.
Note: The chosen database language influences the name of the created database folder.--jdk-version-name
Specifies the JDK version name to be included in the name of the created database folder. This has no effect on the cloned Git repository or the build commands. By default no JDK version name is included in the name of the database folder because it is often not easily possible to obtain the version name only from the specified JDK Git repository.