[clydin]

Modern Yarn formats error detail fields such as Response Code with ANSI color sequences and OSC 8 hyperlinks before passing them to StreamReport, which serializes the formatted strings verbatim into its JSON line output. When colors are enabled, such as by default on GitHub Actions even over non-TTY pipes or via FORCE_COLOR and .yarnrc.yml settings, escape codes appear around both the Response Code label and status number. As a result, parseYarnModernError failed to match the HTTP status code, fell back to the generic YN0035 code instead of E404 for missing packages, and leaked raw control sequences into the parsed summary and detail text.

Strip VT control characters from modern Yarn error messages prior to matching an anchored Response Code line, and ignore 2xx status codes for consistency with the Yarn Classic parser. In addition, split multi-line error payloads such as YN0001 exception stack traces so summary only retains the first line while remaining lines move into detail, fall back to UNKNOWN_ERROR when displayName is an empty string (when enableMessageNames is disabled), and return null directly when no JSON error lines are present instead of delegating to parseNpmLikeError on every successful command output.

[gemini-code-assist]

Code Review

This pull request refactors the modern Yarn error parser to strip ANSI color codes and OSC 8 hyperlinks from error lines, improves HTTP status code extraction, and refines how error summaries and details are split. It also adds comprehensive unit tests to verify these changes. The review feedback suggests a robust fallback for the error summary to prevent it from being empty when the error data contains only whitespace.

[gemini-code-assist]

If first.data is empty or contains only whitespace, the destructured summary will default to an empty string (''). To ensure a robust fallback and prevent empty error summaries, we can default summary to first.code (e.g., 'YN0035' or 'UNKNOWN_ERROR'), which is guaranteed to be a non-empty string.

  const [summary = first.code, ...firstDetailLines] = first.data\n    .split('\\n')\n    .map((line) => line.trim())\n    .filter(Boolean);