Agent Mode in 0.16 crashes with no response and does not create todo plan - It was working in 0.15

[danielerdmannarv]

Feedback source

None

GitHub Copilot plugin version

0.16

Eclipse IDE and version

2026-03 (4.39.0)

Platform

Microsoft Windows NT 10.0.26100.0 x64

Steps to reproduce

Open the GitHub Copilot Chat view in Eclipse.

Provide technical context by opening the following objects:

Function Module: Z_MOP_EWM_MON_CANCEL_POHU_ITM (Legacy)

Function Module: /ACOD/AEWM_MON_CANCEL_POHU_ITM (Template)

Class: ZCL_MOP_EWM_CANCEL_POHU (Target for refactoring)

Class: /ACOD/CL_AEWM_CANCEL_POHU (Base class)

Submit the refactoring prompt provided below.

Here is the prompt we use:

The logic to cancel POHU Item in function module 'Z_MOP_EWM_MON_CANCEL_POHU_ITM' was written in more generic way into function module /ACOD/AEWM_MON_CANCEL_POHU_ITM and class /ACOD/CL_AEWM_CANCEL_POHU. 

Z_MOP* is an implementation for a specific warehouse and /ACOD/ is general for all warehouses.

As Z_MOP logic requires some specific additional logic, class ZCL_MOP_EWM_CANCEL_POHU extends the class /ACOD/CL_AEWM_CANCEL_POHU with this specific logic. This class was already adjusted. You can check it. 

Your Tasks are:
- Implement the class ZCL_MOP_EWM_CANCEL_POHU fully, so that the logic and functions stay as they are in Z_MOP_EWM_MON_CANCEL_POHU_ITM. But as a /ACOD-Inheritance to minimize the logic in Z*
- Change Funtionc Module Z_MOP_EWM_MON_CANCEL_POHU_ITM. It should look like /ACOD/AEWM_MON_CANCEL_POHU_ITM and use class ZCL_MOP_EWM_CANCEL_POHU, similar to how it is done in Function Module /ACOD/AEWM_MON_CANCEL_POHU_ITM.

Lookup  ALV Structure - 'ZMOP_EWM_MON_ALV_POHUITM_CANC' for Z_MOP logic 

This should be a strait forward todo. If you find any issues, that requires a change in /ACOD/, that makes the change easier or is needed, give me a hint.

Logs (optional)

No response

Expected behavior

Screenshot of working 0.15

Upon receiving the prompt, the Copilot agent should perform the following sequence:

• Plan Generation: The agent should analyze the request and generate a list of "Todos" displayed within a collapsible control between the chat input and output.

• Sequential Execution: The agent should systematically work through the generated todo list, applying changes to the relevant ABAP objects (Function Modules and Classes).

• Completion Indicator: The process should conclude with the agent displaying the name of the model used. This serves as the visual confirmation that the task is finished and the chat is ready for the next user prompt.

Actual behavior

Screenshot of broken 0.16

The agent begins the process by reading the files but fails to generate a response or the expected "Todo" list.

• Unresponsive State: The agent remains in a "reading" or "processing" state indefinitely. During testing, the agent was left for 30 minutes without any output or UI updates before we were forced to terminate the process.
• Lack of Planning: No collapsible control or plan was initialized.
• System Impact: The IDE remains functional, but the Copilot agent becomes completely unresponsive to further input.

Additional information

No response