fix: strip ANSI escape sequences before JSONL parsing#24
Open
Mingsicogi wants to merge 3 commits intoopenai:mainfrom
Open
fix: strip ANSI escape sequences before JSONL parsing#24Mingsicogi wants to merge 3 commits intoopenai:mainfrom
Mingsicogi wants to merge 3 commits intoopenai:mainfrom
Conversation
When the Codex CLI runs in environments where the shell emits terminal escape sequences (e.g. bracketed paste mode `[?2004h` from zsh), these control characters leak into the JSONL stream and cause JSON.parse to fail with: "Failed to parse codex app-server JSONL: Unexpected token" Extract a shared `stripAnsi()` helper in `app-server.mjs` and use it in both `app-server.mjs` and `app-server-broker.mjs` to strip ANSI escape sequences from each line before JSON parsing. Fixes openai#23
Mingsicogi
force-pushed
the
fix/strip-ansi-escape-jsonl-parse
branch
from
08103c9 to
665a042
Compare
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 08103c94e1
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
Address review feedback: the initial regex missed CSI sequences with non-letter finals (e.g. `\e[200~` for bracketed paste wrappers) and OSC sequences terminated by ST (`ESC \`) instead of BEL. Updated regex covers: - CSI: ESC [ <params> <final byte 0x40-0x7E> (includes ~, @, etc.) - OSC: ESC ] ... (BEL | ESC \) - Simple escapes: ESC <intermediate>* <final>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 7554408f21
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
Addresses remaining review feedback from chatgpt-codex-connector: CSI (previously missed): - Parameter bytes: [0-9;?] → [\x30-\x3f] to include <, =, > (0x3C–0x3E) e.g. ESC[>4;2m (modifyOtherKeys), ESC[<1;2M (mouse events) - Intermediate bytes: added [\x20-\x2f]* between params and final e.g. ESC[ q (cursor shape), ESC[!p (soft reset) String sequences (new): - DCS (ESC P), SOS (ESC X), PM (ESC ^), APC (ESC _) terminated by BEL or ST e.g. ESC P...ESC\ (tmux/Sixel), ESC _G...ESC\ (Kitty graphics protocol) Simple escapes: - Final byte range: [\x40-\x7e] → [\x30-\x7e] to include Fp sequences (0x30–0x3F) e.g. ESC 7/8 (save/restore cursor), ESC = / ESC > (keypad modes) Lone ESC fallback: - Bare \x1b not matched by any complete sequence is now stripped prevents JSON.parse failures from truncated/unknown sequences Add tests/strip-ansi.test.mjs with 27 cases covering all sequence types. Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
\e[?2004h) from JSONL lines beforeJSON.parse()in bothapp-server.mjsandapp-server-broker.mjsFailed to parse codex app-server JSONL: Unexpected tokenerrors when the shell environment leaks terminal control characters into the Codex CLI output streamProblem
When
/codex:reviewruns in environments where the shell (zsh/bash) emits ANSI escape sequences during startup, these control characters appear in the JSONL stream.JSON.parse()then fails because[?2004his not valid JSON.This affects both foreground and background execution via Claude Code's Bash tool.
Fix
Added a regex-based ANSI escape sequence stripper before JSON parsing in the two JSONL consumers:
plugins/codex/scripts/lib/app-server.mjs—handleLine()methodplugins/codex/scripts/app-server-broker.mjs— socket data handlerLines that become empty after stripping are silently skipped (same as the existing blank-line check).
Test plan
\e[?2004h,\e[0m,\e]0;title\x07and similar sequencesFixes #23