This is a project that bridges the gap between the Java Debugger (JDB) framework available in modern IDEs and native debugging via GNU Project debugger (GDB).

Quarkus is a cloud-native Java development framework. It allows Java code to be mapped to Kubernetes containers and compiled natively. Native compilation is useful for serverless computing, it avoids the overhead of running a JVM in containers and to execute serverless code directly. Today, natively compiled Java code can be debugged using GDB, which is a C/C++ debugger and is unfamiliar to Java developers.

NativeJDB bridges the gap between the Java Debugger Framework available in modern IDEs and native debugging via GDB. It is part of our exploration to improve developer experience and provide familiar tooling for Java cloud-native developers. NativeJDB is useful when the application behaves differently when natively compiled.

NativeJDB uses the JDWP protocol to communicate with an IDE, and acts as a server for it. It wraps a native executable controlled by GDB and makes it possible to control the native executable via JDWP commands. Because of this architecture, NativeJDB can in principle be used with any IDE, but has mostly been tested with IntelliJ. The current implementation of NativeJDB makes use of a "scaffolding VM", which is the same program being executed as a Java process. This allows NativeJDB to obtain certain static informations about the code that are harder to obtain from the GDB. In the future, this scaffolding can be removed.

• Editor opens up automatically after launch/attach with sourcecode showing lines
• Works with familiar IDE like IntelliJ
• Can work with VSCode, and Eclipse with some more exploration and development cycles
• Works with GraalVM's natively compiled images
• Can work with Qbicc's natively compiled images with some more exploration and development cycles
• Debugging features using the IDE's Debug Console itself:
  • Suspend / Resume
  • Set Breakpoints (Insert/Enable)
  • Clear Breakpoints (Delete/Disable)
  • Step Over / Step Into / Step Return
  • Stack Frames
  • Variables (Local + Static) values (Work-in-progress)
  • Thread info

• GraalVM Community Edition for Java 11
• Docker Desktop
• IntelliJ IDE

🎥 https://ibm.box.com/v/nativejdb-demo 🎥

You must install these tools:

1. git: For source control
2. IntelliJ: For debugging native exec via an IDE
3. Docker Desktop: For running NativeJDB in a container

In addition, you must also setup another repository that contains example applications to run via this NativeJDB server:

4. nativejdbExamples: Follow instructions in its README to generate a native image executable before proceeding

To build the NativeJDB, run:

make compile

This produces a jar file NativeJDB-1.0-SNAPSHOT.jar under /target which is used in Docker container.

To generate a native executable within the Linux environment in the Docker container, you will need to download graalvm binary from https://github.com/graalvm/graalvm-ce-builds/releases and untar graalvm-ce-java11-linux-amd64-*.tar.gz into the same NativeJDB directory.

 make graalvm

• Start Docker Desktop

• Run the following command via a terminal to deploy docker container running NativeJDB server to start debugging:

For an already pre-built native image using nativejdbExamples (like Hello), run this:

make nativejdb CLASSNAME=Hello 

For any other application that has your own pre-built native executable and debug sources, add the executable and debug sources to /apps directory and then pass the following input args to make target:

make nativejdb CLASSNAME={nameofjarfile} NATIVEEXEC=apps/{nameofnativeexec} NATIVESRC=apps/{directorynameofdebugsources}

This will start nativejdb in a running Docker container.

To connect an IDE debugger to NativeJDB, the user needs to create a Debug configuration and attach to NativeJDB on the appropriate port.

Set breakpoints in the source code file for your example application in the nativejdbExamples project (for example: nativejdbExamples/src/Hello/Hello.java

On IntelliJ, from the nativeJDBExamples open project: Run ---> Edit Configurations --> Remote JVM Debug --> nativejdbExamples/.run/Hello

• Single threaded programs only
• Short running programs need a thread sleep (so program does not end before NativeJDB has time to start a scaffolding VM and attach to it)
• Hot Code Replace not possible
• Known issue: breakpoints in loops work only once (related to this graalvm issue)
• Known issue: step operation sometimes continues instead

If you are interested in contributing, see CONTRIBUTING.md for contributing information

https://quarkus.io/guides/building-native-image#debugging-native-executable

https://sourceware.org/gdb/onlinedocs/gdb/GDB_002fMI.html

To discuss more about this project, please open an issue.

This project consists of a frontend package with a GPL-EC license (code derived from: https://github.com/JetBrains/jdk-sa-jdwp) and a backend package with an EPLv2 license (code derived from: https://github.com/eclipse-cdt/cdt). The top-level project has an MIT license.