fix: handle null response.output in parse_response

[C1-BA-B1-F3]

Summary

parse_response() in src/openai/lib/_parsing/_responses.py iterates response.output without checking for None. The chatgpt.com Codex backend (https://chatgpt.com/backend-api/codex) sometimes sends response.output: null in the consolidated response.completed event — even when valid response.output_item.done events were streamed earlier in the same response. The SDK then raises TypeError: 'NoneType' object is not iterable inside the stream accumulator, killing the entire stream before the consumer can read the deltas.

Fix

One-line defensive iteration in parse_response:

-    for output in response.output:
+    for output in (response.output or []):

With this patch the stream completes, get_final_response() returns a Response with output=[], and consumers that already track output_item.done events can backfill the response from their own collected items. No regression on responses that legitimately have output: list.

Tests

Added tests/lib/test_parsing_responses.py with three tests:

• test_parse_response_with_none_output — verifies null output doesn't crash and returns empty output list
• test_parse_response_with_empty_list_output — verifies empty list still works
• test_parse_response_with_message_output — verifies normal output items still parse correctly

Fixes #3325

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d91d0d75c2

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".

[chatgpt-codex-connector]

Preserve streamed output for null completed responses

When this runs from ResponseStreamState.accumulate_event() for a response.completed event, event.response.output == None is now converted to an empty final ParsedResponse.output even though the stream state may already contain items accumulated from prior response.output_item.added / text delta events. In that null-output backend scenario, consumers of response.completed.response, get_final_response().output, or get_final_response().output_text silently get an empty response instead of the text/tool calls they just streamed, so the stream no longer crashes but still drops the completed result.

Useful? React with 👍 / 👎.

[C1-BA-B1-F3]

Thanks for raising this! You're correct that a null in was overwriting the accumulated stream content, resulting in empty final responses.

The fix preserves accumulated items from prior / text delta events when is null, so consumers get the complete streamed result rather than an empty response. The null output case now only applies when there's genuinely nothing streamed.

Happy to adjust the approach if you prefer a different strategy for handling this edge case!