-
Notifications
You must be signed in to change notification settings - Fork 1.7k
Open
Labels
Feature: Copilot HoverGitHub Copilot hover "Generate Copilot summary" featureGitHub Copilot hover "Generate Copilot summary" featureLanguage ServiceVisual Studio CodeCaused by (or depends on changes from) VS CodeCaused by (or depends on changes from) VS Codebug
Description
Environment
- VS Code Remote: WSL
- Host OS: Windows + WSL2 (Ubuntu)
- VS Code Server commit:
0870c2a0c7c0564e7631bfed2675573a94ba4455 - C/C++ extension version:
ms-vscode.cpptools 1.30.5 (linux-x64) - Workspace type: mixed monorepo (not primarily C/C++)
Bug Summary and Steps to Reproduce
Problem
After updating, VS Code extension host CPU became abnormally high.
The issue appears tied to cpptools repeatedly trying to resolve language models through LanguageModelProxy.
Observed behavior:
extensionHostprocess CPU stays elevated (continuous activity, no idle)- cpptools-related processes are active
- logs are flooded with repeated warnings:
LanguageModelProxy](ms-vscode.cpptools) Could not find model 'copilot/...'. Trying to resolve model again.
Evidence
From remoteexthost.log:
- Repeated lines like:
Could not find model 'copilot/claude-haiku-4.5' in local cache. Trying to resolve model again.Could not find model 'copilot/gpt-4.1' in local cache. Trying to resolve model again.Could not find model 'copilot/gpt-4o' in local cache. Trying to resolve model again.Could not find model 'copilot/auto' in local cache. Trying to resolve model again.
Frequency observed:
- ~30k+ total repeated warnings in one log session
- ~1.9k warnings in a short recent tail sample
Repro (what happened)
- Update VS Code / Remote server.
- Open Remote-WSL workspace.
- cpptools activates.
- extension host starts high-frequency model-resolution retries (warning spam + CPU increase).
Temporary Mitigation
Disabling cpptools in remote WSL immediately reduced CPU usage.
(renamed extension dir to disable and restarted extension host)
Expected
cpptools should not continuously retry model resolution in a tight loop when models are unavailable/unresolvable.
It should back off aggressively or stop retrying after bounded attempts.
Additional note
This looks like a retry/backoff regression or integration mismatch between cpptools LanguageModelProxy usage and current model cache/availability in Remote-WSL.
Configuration and Logs
/root/.vscode-server/extensions/ms-vscode.cpptools-1.30.5-linux-x64.disabled-20260305-174449
I disabled itOther Extensions
No response
Additional context
No response
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
Feature: Copilot HoverGitHub Copilot hover "Generate Copilot summary" featureGitHub Copilot hover "Generate Copilot summary" featureLanguage ServiceVisual Studio CodeCaused by (or depends on changes from) VS CodeCaused by (or depends on changes from) VS Codebug
Type
Projects
Status
No status