openai / openai/codex

[Windows][26.908.4834.0] Renderer C000001D at chrome.dll+0xAD95EE0; identical raw_hash_map::at fatal in two dumps

Open
#45,247 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

app bug windows-os
Dominant language
Rust
Stars
125k
Forks
19.4k
PR merge metrics
PR metrics pending

Description

Summary

On Windows, Codex Desktop's renderer is terminating through the same native fatal path in two independently captured dumps. This persists on desktop package 26.908.4834.0.

This is more specific than my earlier idle-startup report (#44115), which did not have native dumps. I am filing it separately so the renderer failure can be symbolized and triaged without conflating it with browser-main crashes.

Environment

  • Windows 11 24H2, x64
  • Codex Desktop package: OpenAI.Codex 26.908.4834.0
  • Embedded Chromium: 152.0.7977.83

Confirmed native failure (two captures)

Two independent ProcDump MiniPlus captures of processes launched with --type=renderer have the same:

  • unhandled exception: C000001D (ILLEGAL_INSTRUCTION)
  • fault instruction: chrome.dll+0xAD95EE0, disassembling to ud2
  • WER/CDB bucket: ILLEGAL_INSTRUCTION_c000001d_chrome.dll!Unknown
  • Failure.Hash: {f25bb7c2-d945-c864-f5f8-519ce06a7b76}
  • fatal text present in each dump: throw_delegate.cc:107 and raw_hash_map<>::at

For the 2026-09-13 19:05 UTC+08:00 capture, ProcDump reported the renderer's unhandled C000001D, completed the dump, and Crashpad concurrently recorded ptype=renderer.

The browser-main process and codex.exe app-server survived; a replacement renderer was created immediately afterward. This is therefore a renderer-only native termination, not a full desktop-app exit.

Observed state around reload (correlation only)

On some renderer reloads, the UI recovery path also logs Conversation state not found and Item not found in turn state, particularly while rehydrating a long active task. I am not claiming those messages are the direct native cause; the exact same renderer fatal path was confirmed only by the two dumps above.

Request

Could the team symbolize chrome.dll+0xAD95EE0 for this package and investigate the raw_hash_map<>::at fatal/missing-key path in the Codex renderer? The two matching dumps strongly suggest a deterministic product code path rather than a one-off system fault.

No conversation content, local paths, credentials, full logs, or dump files are attached publicly. I can provide the two dumps and their SHA-256 values through a private/support-approved channel if useful.

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.

Research direction

Start by symbolizing chrome.dll+0xAD95EE0 for package 26.908.4834.0 and reviewing the two ProcDump MiniPlus captures through the private channel. Trace the raw_hash_map<>::at fatal path in the renderer; done means identifying the missing-key call path and a reproducible product-level cause.

Written by the indexing model from the issue text.

Assessment

Tech stack
rust
Domain
desktop, operating-systems
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Active
Clarity
Mostly clear
Newbie friendliness
28/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.