ENOENT: no such file or directory, rename â/Users/perso/.gemini/antigravity/user_settings.pb.tmp.extHostâ â â/Users/perso/.gemini/antigravity/user_settings.pbâ
Messages (4)
Menu item references a command antigravity.importAntigravitySettings which is not defined in the âcommandsâ section.
Antigravity Version: 1.13.3
VSCode OSS Version: 1.104.0
Commit: 94f91bc110994badc7c086033db813077a5226af
Date: 2025-12-19T21:03:14.401Z (3 wks ago)
Electron: 37.3.1
Chromium: 138.0.7204.235
Node.js: 22.18.0
V8: 13.8.258.31-electron.0
OS: Darwin arm64 25.1.0
Language Server CL: 846830895
I lose a lot of time with these issues. Uninstall, delete all preferences, install again.
Subject: EMERGENCY WORKAROUND: Root Cause and Fix for v1.13.3 âunleashWrapper is nilâ Error
Hi everyone,
After coordinating with my team and analyzing the logs (specifically from @dev_noersy and @Milad_Alem), we have identified the most probable issue. This is a Level 5 Code Regression in version 1.13.3 affecting all architectures (macOS Tahoe, Windows 11, and Linux).
The Diagnosis:
The error âunleashWrapper is nilâ indicates a race condition. The extension is trying to check your token limits (via the Unleash feature-flag client) BEFORE the Authentication handshake is finished. Because the wrapper hasnât initialized yet, it hits a null pointer and kills the LLM stream.
Tactical Workarounds (Do these until a hotfix is pushed):
1. The â60-Second Cold Startâ: Launch Antigravity and DO NOT click âGenerate Commitâ immediately. Wait at least 60 seconds to allow the background Auth and Feature-Flag threads to sync. This bypasses the race condition on most systems.
2. The Settings Flush (For macOS/Linux): If you see the ENOENT rename error, your settings file is stuck in a temp state. Close the app and run:
If you are blocked, use the Gemini CLI or standard git commit in the terminal for now. The LLM logic in the IDE is currently gated by this broken initialization wrapper.
Message for Google Devs:
The LanguageServerService needs an initialization guard on the unleashWrapper. It should gracefully degrade to a default token limit if the wrapper is nil instead of throwing a stream-killing error.
Thanks for the info. I did wait a lot before but it did not help. Iâll test other workarounds but honestly, Iâve switched to asking the agent itself to commit (during the conversation) and itâs better because it adds descriptions to the commits as, well. The generate button only generates a message.
Iâm wondering how itâs a race condition on Auth if the agent itself is working fine and is clearly passed the auth part?
Excellent question, @Milad_Alem. Here is why Chat works while Commits fail:
1. Service Decoupling: In Antigravity, the âAgent/Chatâ and the âCommit Generatorâ likely use different gRPC interceptors. The Chat service may be checking Auth at the edge, while the Commit service has an additional âgatekeeperâ (the unleashWrapper) that specifically checks for token quotas before it even talks to the LLM.
2. The Null Pointer: Even if you are âAuthed,â if the specific internal object responsible for counting the tokens (unleashWrapper) failed to instantiate during the boot sequence, any feature that calls it (like Commit Gen) will crash. Chat might simply not be calling that specific wrapper yet, or it has a different fallback mechanism.
3. Async Initialization: Itâs a âRace Conditionâ not of the Auth itself, but of the internal state. The Auth might be âGreen,â but the Feature-Flag client is still âGray.â If you trigger a commit while itâs in that âGrayâ state, it hits a nil reference.
Think of it like a starship: The main reactor (Auth) is online so the lights (Chat) are on, but the targeting computer (Unleash Wrapper) failed its boot sequence. You can talk to the crew, but you canât fire the main phasers.