App crashes with V8 OOM in "git" worker thread during long computer-use sessions

Open
#34,213 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
45/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Quiet
Tech stack
git, javascript, node.js
Domain
desktop, tooling

Research direction

Start with the reported V8 crash stack and the Node worker thread named "git"; request or inspect the available .ips crash logs and review how that worker is used during computer-use/testing sessions. Reproduce a long session, then verify that the app remains stable without the git worker exhausting the V8 heap.

Written by the indexing model from the issue text.

Description

app bug computer-use performance
What version of the Codex App are you using (From “About Codex” dialog)?

26.715.31925 (build 5551)

What subscription do you have?

Plus

What platform is your computer?

Darwin 25.5.0 arm64 arm

What issue are you seeing?

App repeatedly crashes during long, goal-directed agentic tasks using the computer-use/testing feature. Happened 3 times so far, at roughly 87min, 71min, and 28min into separate sessions — duration varies but the pattern is consistent.

Root cause (confirmed from crash logs, not a guess): a V8 JavaScript heap out-of-memory in a Node worker thread named "git" that Codex spins up during these sessions. Node's own OOM handler catches it and calls abort():

abort()
→ node::OOMErrorHandler(char const*, v8::OOMDetails const&)
→ v8::ExternalMemoryAccounter::Update(...)

This looks like a memory leak in the git-integration worker that grows across a long session rather than a fixed timeout — the crash happens at different elapsed times but always in the same subsystem. One occurrence crashed while the app was backgrounded (not in focus), so it's not tied to active interaction either.

An earlier crash from the same setup (before this one) showed the same subsystem faulting via EXC_BREAKPOINT instead of SIGABRT — same root cause, different exit path, consistent with memory pressure building up differently each time.

In-app feedback already submitted, Feedback ID: no-active-thread-019f7ba0-0cb3-7282-80aa-966f2ddc7cdf

Full .ips crash logs available for both incidents, happy to attach.

What steps can reproduce the bug?

Feedback ID: no-active-thread-019f7ba0-0cb3-7282-80aa-966f2ddc7cdf

What is the expected behavior?

The Codex app should remain stable during long computer-use/testing sessions, and the Git integration worker should release or bound memory instead of exhausting the V8 heap and terminating the app.

Additional information

No response

Dominant language
Rust
Stars
125k
Forks
19.5k
Avg merge
1m
Merged PRs (30d)
1k

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from openai/codex

All issues in openai/codex

Similar issues

More Rust issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.