Skip to content
This repository was archived by the owner on May 15, 2026. It is now read-only.
This repository was archived by the owner on May 15, 2026. It is now read-only.

Extension crashes VS Code via repeated deletePointsByMultipleFilePaths failures #5516

Closed

Description

@buck-0x

App Version

3.23.3

API Provider

Anthropic

Model Used

N/A

Roo Code Task Links (Optional)

The roo-cline extension is triggering a cascade of unhandled errors that destabilize the VS Code Extension Host and ultimately crash the editor.

##Reproduction context:

Version: 1.101.2
Commit: 2901c5ac6db8a986a5666c3af51ff804d05af0d4
Date: 2025-06-24T20:27:15.391Z
Electron: 35.5.1
ElectronBuildId: 11727614
Chromium: 134.0.6998.205
Node.js: 22.15.1
V8: 13.4.114.21-electron.0
OS: Darwin arm64 24.5.0

Project:

  • Workspace: Monorepo with active file watchers

Observed behavior:

The extension enters a tight error loop on startup:

Failed to delete points by file paths: Error: Bad Request
at deletePoints
at deletePointsByMultipleFilePaths
at processBatch
  • Repeats dozens of times in under a second.
  • Floods the logs and overwhelms the extension host.
  • Triggers VS Code crash (verified via system and VS Code logs).

🔁 Steps to Reproduce

1.	Install the rooveterinaryinc.roo-cline extension.
2.	Open a VS Code workspace that includes a directory with many files (e.g. a TypeScript monorepo).
3.	Ensure the extension is enabled and VS Code’s telemetry sync is active.
4.	Start VS Code and observe:
•	Log output (Cmd+Shift+U → Output → Log (Extension Host))
•	CPU spikes in Activity Monitor (macOS) or top (CLI)
5.	Within seconds, the errors appear repeatedly
6.     VS Code slows down and crashes or hangs due to extension host overload.

💥 Outcome Summary

No errors in the VS Code developer console and no crashses

📄 Relevant Logs or Errors (Optional)

Activity

  1. dosubot commented on Jul 9, 2025

    @dosubot

    Hi @buck-0x! I'm Dosu and I’m helping the Roo-Code team.

    I couldn't find any documentation or known issues about deletePointsByMultipleFilePaths or repeated 'Bad Request' errors causing extension host crashes. There is, however, guidance for handling large projects: try lowering the "File read auto-truncate threshold" in Roo Code's Advanced Settings to reduce resource usage, especially in large workspaces or monorepos. This setting limits how much of each file is read at once and can help prevent overloads in big projects. If the issue persists, you might also try reducing the number of files watched or temporarily disabling telemetry sync to see if that changes the error pattern. There’s a documentation gap here—no official troubleshooting steps for this specific error loop or crash scenario were found.
    Tips for large projects and resource usage
    Managing large files and performance

    To reply, just mention @dosu.


    Docs are dead. Just use Dosu.

    Leave Feedback Ask Dosu about Roo-Code Share Dosu with your team

  2. added
    Issue/PR - TriageNew issue. Needs quick review to confirm validity and assign labels.
    on Jul 9, 2025
  3. moved this from Triage to Issue [Needs Scoping] in Roo Code Roadmapon Jul 9, 2025
  4. daniel-lxs commented on Jul 9, 2025

    @daniel-lxs
    Member

    Hey, thank you for reporting this. I took a look and it seems the crash loop is caused by the file watcher's batch processing logic, specifically during calls to deletePointsByMultipleFilePaths in the Qdrant vector store. This triggers repeated "Bad Request" errors that eventually crash the VS Code extension host.

    The issue starts when the extension processes a large number of file changes. It tries to delete existing vector embeddings before inserting new ones, but the Qdrant API rejects the delete request, likely due to malformed filters or request size limits. Since the error is re-thrown with no retry logic or backoff, it creates a tight loop that floods the logs and eats CPU until the extension host crashes.

    This typically happens on startup because the indexing service runs automatically when the extension activates. In large projects, that immediately kicks off batch processing, which hits the bug.

    To fix this, we probably need to:

    • Add backoff and retry logic for failed Qdrant operations
    • Limit the number of file paths per request
    • Handle failures more gracefully so they don't crash the host
  5. moved this from Issue [Needs Scoping] to Issue [Unassigned] in Roo Code Roadmapon Jul 9, 2025
  6. added and removed
    Issue/PR - TriageNew issue. Needs quick review to confirm validity and assign labels.
    on Jul 9, 2025
  7. self-assigned this
    on Jul 9, 2025
  8. moved this from Issue [Unassigned] to Issue [In Progress] in Roo Code Roadmapon Jul 9, 2025
  9. added
    Issue - In ProgressSomeone is actively working on this. Should link to a PR soon.
    and removed on Jul 9, 2025
  10. niteshbalusu11 commented on Jul 14, 2025

    @niteshbalusu11
    Contributor

    I am hitting this issue on some of my codebases.

  11. moved this from Issue [In Progress] to Done in Roo Code Roadmapon Jul 16, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

Issue - In ProgressSomeone is actively working on this. Should link to a PR soon.bugSomething isn't working

Type

No type

Projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions