[Bug][ghidra2cpg] Ghidra frontend does not support indirect jumps

[Ik0ri4n]

Describe the bug
The Ghidra frontend does not resolve indirect jumps or use Ghidra's analysis for that. Instead, the resulting control flow will simply miss the corresponding edges. The resulting CFG will contain disconnected branches.

This is especially relevant since indirect jumps are frequently used to implement switch-case-expressions. Any flow analysis on such a program will be incomplete.

To Reproduce
Steps to reproduce the behavior:

1. (Optional) Recompile with your compiler of choice
   • You need to verify that the program actually uses a jump table!
   • You may need to modify the program to force that!
2. Import the binary: importCode("../indirect-jump/test", "broken-jumps", language="GHIDRA")
3. Inspect the generated CFG: cpg.method("main").plotDotCfg

Expected behavior
The CFG contains edges for each case branch, if it can be recovered by Ghidra.

Screenshots

Desktop

• OS Ubuntu 25.04
• Joern Version 4.0.468
• Java Version 21.0.9
• GCC Version gcc (Ubuntu 14.2.0-19ubuntu2) 14.2.0 (used to compile the provided binary)

Attachments
test.zip
test.c

[Ik0ri4n]

This clearly corresponds to the missing implementation in JumpPass.

Considering the proclaimed "Maturity: High" of the Ghidra Frontend, I'd consider this a bug or at least relevant enough to be tracked in an issue.

For future reference: From a user perspective, a warning would be extremely useful in any such case. It would document the issue and prevent the issue going unnoticed.

[Ik0ri4n]

During the analysis of this bug and debugging of other errors with generated CPGs, I found two other bugs relevant for the correct and complete generation of CFGs. I believe that these bugs are important and may also be caused by structural issues in the code base.
I wanted to ask beforehand, whether you are interested in a discussion of these issues and what would be the best way to do so? I'd like to avoid creating multiple issues for similar topics.

Also, let me state in advance that I have a lot on my schedule until early June. I would like to help address the issues I found and potentially upstream fixes I implemented, if you are interested in that. Those fixes are temporary fixes though and would likely be better addressed by partial rewrites of the frontend, at least in my opinion. This would require a time commitment I cannot currently make.

Finally, please don't take the issues I described as personal critique. I believe that the Ghidra frontend is already quite powerful and am thankful that it exists. In fact, I did use it quite extensively recently (hence how I encountered these bugs) and it was a great help. Based on my experience, I do however believe that its stability and support for general binaries can be improved.