error: zsh: suspended (tty input) codex --yolo --search resume

ah maybe a fix will land here?: https://github.com/openai/codex/pull/9403

Do you have any non-default configuration settings in your config.toml? Perhaps some experimental features enabled?

No experimental features enabled.

Not that much there:

```config.toml
model = "gpt-5.2"
model_reasoning_effort = "xhigh"
notify = ["bash", "-lc", "/usr/bin/afplay /System/Library/Sounds/Bottle.aiff"]

tool_output_token_limit = 25000

<various trust_level for project settings />

[notice]
"hide_gpt-5.1-codex-max_migration_prompt" = true

[notice.model_migrations]
"gpt-5.1-codex-max" = "gpt-5.2-codex"
"gpt-5.2" = "gpt-5.2-codex"

[mcp_servers.code-mode]
command = "/Users/matt/dotfiles/scripts/code-mode-bridge"

[mcp_servers.RepoPrompt]
command = "/Users/matt/RepoPrompt/repoprompt_cli"
args = []
tool_timeout_sec = 10000
enabled = false

latest round of vibed recursive reasoning before I ran out of tokens, may have found a mitigation:

─ Worked for 18m 55s ────────────────────────────────────────────────────────────────────────────────────────────────────────

• Yep — there’s a very consistent pattern now:

• Every “kill” in the transcripts is actually zsh job-control stopping Codex with SIGTTIN, i.e. zsh: suspended (tty input)

codex --yolo --search resume. That’s not an OOM/segfault; it’s “a process tried to read from the controlling TTY while
not the foreground pgrp”.

• It always happens while Codex is in a long-running state (“esc to interrupt”), which strongly suggests Codex (or a non-

PTY child it spawns) is touching /dev/tty at the wrong time. That matches https://github.com/openai/codex/pull/9403
exactly: it adds an experimental detach_non_tty feature that does setsid() for non-PTY child spawns so they can’t
inherit /dev/tty (and therefore can’t trigger these SIGTTIN stops).

ASAP mitigation already landed in your dotfiles

• zsh/path/codex:5 wraps Codex and ignores SIGTTIN (trap '' TTIN) so it shouldn’t get job-stopped anymore.
• zsh/lib/prompt.zsh:149 + zsh/lib/tty-sanitize.zsh:1 auto-restores stty + clears stuck terminal modes before every prompt,

so even if Codex dies/stops, your shell stops “spewing garbage”.

• scripts/codex-tty-doctor:1 gives you codex-tty-doctor fix / report for emergency recovery + debugging.

Next thing to do (important): restart your shell + restart Codex so the wrapper actually applies (it won’t affect an
already-running codex --yolo --search resume). Then verify which codex resolves to ~/dotfiles/zsh/path/codex.

Approaching rate limits
Switch to gpt-5.1-codex-mini for lower credit usage?

› 1. Switch to gpt-5.1-codex-mini Optimized for codex. Cheaper, faster, but less capable.

1. Keep current model
2. Keep current model (never show again) Hide future rate limit reminders about switching models.

i ran out of tokens, and it's bedtime, so i'll have to see if it continues to occur tomorrow.

Thanks for the bug report. We think we understand the problem and are working on a fix.

This is happening repeatedly on all my codex sessions. I resume a session, get one prompt, and if I dare type anything in the input, it crashes, even midstream. This is messed up.