claude-code - 💡(How to fix) Fix Orphaned Claude session survives 11h after worktree deletion, ignores SIGTERM [1 comments, 2 participants]

TL;DR

The Claude Code process may need a modification to handle the case when its working directory is deleted, such as periodically checking for the directory's existence and exiting if it no longer exists.

Guidance

• Investigate the Claude Code source code to determine if there is a mechanism for periodically checking the existence of the working directory and handle the case when it is deleted.
• Consider adding a signal handler for SIGTERM that checks if the working directory still exists before attempting to exit, to ensure the process can be terminated cleanly.
• Review the system's process management configuration to understand why the process ignored SIGTERM and required SIGKILL to terminate.
• Check if there are any open issues or pull requests in the Claude Code repository related to this problem, as it may have already been addressed or be in the process of being fixed.

Example

No specific code example can be provided without access to the Claude Code source code, but a general approach might involve adding a periodic check for the working directory's existence, such as:

import os
import time

while True:
    if not os.path.exists('/path/to/working/directory'):
        # Handle the case when the directory no longer exists
        break
    # Continue with normal operation
    time.sleep(60)  # Check every minute

Notes

The exact solution will depend on the specifics of the Claude Code implementation and its requirements. This issue may be related to how the process handles file system events or signal handling.

Recommendation

Apply a workaround, such as implementing a periodic check for the working directory's existence, as the root cause of the issue is likely related to the process's handling of its working directory and signal handling, which may require modifications to the Claude Code source code.