"Failed to persist policy" error despite correct permissions · google-gemini/gemini-cli#19919

(15 留言) (0 反應) (1 負責人)TypeScript (103,992 star) (13,657 fork)batch import

area/coreeffort/smallhelp wantedkind/bugpriority/p2status/bot-triagedtype/bug

描述

What happened?

I am receiving the error "✕ Failed to persist policy for run_shell_command" when selecting "allow for all future sessions". I have verified the following: 1. Permissions: The directory ~/.gemini/policies is writable by my user. 2. Configuration: settings.json has "enablePermanentToolApproval": true. 3. File State: The auto-saved.toml file exists, is valid TOML, and is not locked by other processes. This appears to be an internal issue with the CLI's persistence logic or sandboxing constraints preventing the write operation. System: Linux (aarch64)

[ACTION REQUIRED] 📎 PLEASE ATTACH THE EXPORTED CHAT HISTORY JSON FILE TO THIS ISSUE IF YOU FEEL COMFORTABLE SHARING IT.

What did you expect to happen?

preference stored

Client information

CLI Version: 0.29.5
Git Commit: 2ef872e73
Session ID: 824fa46c-018a-4ba6-b765-1f305fd73dc6
Operating System: linux v22.21.1
Sandbox Environment: no sandbox
Model Version: gemini-3-pro-preview
Auth Type: oauth-personal
Memory Usage: 1.04 GB
Terminal Name: Unknown
Terminal Background: #0c0c0c
Kitty Keyboard Protocol: Unsupported

Login information

No response

Anything else we need to know?

No response

貢獻者指南

技術棧: typescriptnodejs
領域: cli
議題類型: bug
難度: 3
預計時間: half day
活動狀態: blocked
清晰度: mostly clear
前置要求: basic understanding of CLI toolsfamiliarity with TypeScriptknowledge of file I/O in Node.js
新手友善度: 30
研究方向: Investigate the policy persistence logic in the gemini cli codebase, likely in modules related to tool approval and configuration. Look for the function that saves policies (e.g., persistPolicy) and trace the error message 'Failed to persist policy for run shell command'. Check the file handling of ~/.gemini/policies directory and the auto saved.toml file. Consider sandboxing constraints if any. Review the 15 comments in the issue for additional context and potential debugging steps.