Codex App crashes when in-app browser opens a site with failed TLS / fake-ip routing
What version of the Codex App are you using (From “About Codex” dialog)?
OpenAI.Codex_26.707.9981.0_x64
What subscription do you have?
Plus
What platform is your computer?
Microsoft Windows NT 10.0.22631.0 x64
What issue are you seeing?
The Codex Windows desktop app crashes when I use the in-app browser to open:
https://jiaowu.hufe.edu.cn/sso.jsp
The website itself may be unreachable from my current network environment, but the Codex app should show a recoverable browser/network error instead of crashing the whole desktop app.
Observed evidence:
- Windows Event Viewer records a MoAppCrash for ChatGPT.exe / OpenAI.Codex.
- Exception code: 0xc06d007f.
- Faulting module: unknown.
- Codex logs include in-app browser / WebView lifecycle messages followed by “No target available”.
- Browser navigation returned ERR_CONNECTION_CLOSED.
- curl HTTPS testing failed during TLS handshake.
- Local DNS resolves jiaowu.hufe.edu.cn to 198.18.0.8, likely a mihomo fake-ip address.
- A DoH lookup showed the real IP as 59.71.96.15.
What steps can reproduce the bug?
- Use Windows with mihomo / Meta Tunnel / TUN / fake-ip enabled.
- Open the Codex Windows desktop app.
- Use the Codex in-app browser / Browser Use.
- Navigate to:
https://jiaowu.hufe.edu.cn/sso.jsp
- Observe that the page does not load normally and the Codex desktop app may crash.
Notes:
- The website may require a specific campus network or VPN, so the network failure itself may be expected.
- The bug is that Codex crashes instead of showing a recoverable browser/network error.
What is the expected behavior?
Even if the website connection fails, the TLS handshake fails, or DNS/fake-ip routing is abnormal, Codex should only show a page loading failure or network error.
The in-app browser tab may fail or become stale, but the whole Codex desktop app should not crash.
Additional information
This appears to be a Codex in-app browser stability issue triggered by a specific network environment involving mihomo TUN/fake-ip routing and a failed HTTPS connection.
Suggested investigation areas:
- Whether the in-app browser WebView target disappears during navigation failure.
- Whether Browser Use / CDP / Playwright commands are being sent after the browser target is already unavailable.
- Whether ERR_CONNECTION_CLOSED, TLS handshake failure, or empty server replies can trigger an unhandled exception.
- Whether WebView or browser target failures are properly isolated from the main Codex desktop process.
I can provide attachments containing Windows Event Viewer excerpts, Codex log snippets, and network [diagnostics.
This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗