Omit undersized WAV output from Code Mode (#41934)

## Why

Audio models cannot reliably encode tool-generated clips shorter than 25 ms.

## What changed

- Measure the available audio frames in base64-encoded PCM and IEEE float WAV outputs.
- Replace clips under 25 ms with explanatory text while preserving surrounding Code Mode output.
- Leave clips at or above the threshold and unrecognized audio formats unchanged.

## Testing

- Cover the duration boundary across multiple sample rates and all supported `audio()` input forms.
- Verify bounded and truncated WAV chunks, output after `yield_control()`, and end-to-end mixed text and audio output.

GitOrigin-RevId: 342b4bdbc65db4238580f9cdc372763c2576eca5

codex-rs/core/tests/suite/code_mode.rs

@@ -38,6 +38,7 @@ use codex_protocol::models::ContentItem;
 use codex_protocol::models::ImageDetail;
 use codex_protocol::models::PermissionProfile;
 use codex_protocol::models::ResponseItem;
+use codex_protocol::openai_models::InputModality;
 use codex_protocol::openai_models::ToolMode;
 use codex_protocol::protocol::AskForApproval;
 use codex_protocol::protocol::EventMsg;
@@ -3856,6 +3857,89 @@ text("after");
     Ok(())
 }
 
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn code_mode_omits_short_audio_and_preserves_other_output() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let [short_audio_url, valid_audio_url] = [120_u32, 600].map(|frames| {
+        let sample_rate = 24_000_u32;
+        let data_size = frames * 2;
+        let mut wav = b"RIFF".to_vec();
+        wav.extend_from_slice(&(36 + data_size).to_le_bytes());
+        wav.extend_from_slice(b"WAVEfmt ");
+        wav.extend_from_slice(&16_u32.to_le_bytes());
+        wav.extend_from_slice(&1_u16.to_le_bytes());
+        wav.extend_from_slice(&1_u16.to_le_bytes());
+        wav.extend_from_slice(&sample_rate.to_le_bytes());
+        wav.extend_from_slice(&(sample_rate * 2).to_le_bytes());
+        wav.extend_from_slice(&2_u16.to_le_bytes());
+        wav.extend_from_slice(&16_u16.to_le_bytes());
+        wav.extend_from_slice(b"data");
+        wav.extend_from_slice(&data_size.to_le_bytes());
+        wav.resize(44 + data_size as usize, /*value*/ 0);
+        format!("data:audio/wav;base64,{}", BASE64_STANDARD.encode(wav))
+    });
+    let code = format!(
+        r#"
+text("before");
+audio({short_audio_url:?});
+text("between");
+audio({valid_audio_url:?});
+text("after");
+"#,
+    );
+
+    let server = responses::start_mock_server().await;
+    let mut builder = test_codex()
+        .with_model_info_override("gpt-5.4", |model| {
+            model.input_modalities.push(InputModality::Audio);
+        })
+        .with_config(|config| {
+            let _ = config.features.enable(Feature::CodeMode);
+        });
+    let test = builder.build_with_auto_env(&server).await?;
+    responses::mount_sse_once(
+        &server,
+        sse(vec![
+            ev_response_created("resp-1"),
+            ev_custom_tool_call("call-1", "exec", &code),
+            ev_completed("resp-1"),
+        ]),
+    )
+    .await;
+    let second_mock = responses::mount_sse_once(
+        &server,
+        sse(vec![
+            ev_assistant_message("msg-1", "done"),
+            ev_completed("resp-2"),
+        ]),
+    )
+    .await;
+
+    test.submit_turn("use exec to return short and valid audio with text")
+        .await?;
+
+    let req = second_mock.single_request();
+    let items = custom_tool_output_items(&req, "call-1");
+    let (_, success) = custom_tool_output_body_and_success(&req, "call-1");
+    assert_ne!(success, Some(false));
+    assert_eq!(
+        &items[1..],
+        &[
+            serde_json::json!({ "type": "input_text", "text": "before" }),
+            serde_json::json!({
+                "type": "input_text",
+                "text": "Audio output omitted because the clip is shorter than 25 ms; use a longer clip."
+            }),
+            serde_json::json!({ "type": "input_text", "text": "between" }),
+            serde_json::json!({ "type": "input_audio", "audio_url": valid_audio_url }),
+            serde_json::json!({ "type": "input_text", "text": "after" }),
+        ]
+    );
+
+    Ok(())
+}
+
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn code_mode_replaces_malformed_image() -> Result<()> {
     skip_if_no_network!(Ok(()));