fix: 完善测试套件

test/okvm-test/README.md

@@ -76,8 +76,8 @@ python okvm_testctl.py run `
 
 控制端会先通过 SSH 执行 `lsusb -t`，再结合 `/api/devices` 自动选择视频输入：
 
-- USB2.0 采集卡：测试 `1080p30 MJPEG`，再切换 `1080p YUYV` 并选择该分辨率最高帧率；如果没有 1080p YUYV，才退到不超过 1080p 的最高分辨率。
+- USB2.0 采集卡（MS2131 测试档位）：严格测试 `1080p50 MJPEG` 和 `1080p10 YUYV`；任一格式、分辨率或帧率未申报时，立即将 `video_input_select` 标记为 `FAIL` 并中止测试，不回退到其他档位。
-- USB3.0 采集卡：测试 `1080p60 MJPEG`，再切换 `1080p YUYV` 并选择该分辨率最高帧率；如果没有 1080p YUYV，才退到不超过 1080p 的最高分辨率。
+- USB3.0 采集卡：测试 `1080p60 MJPEG`，再切换 `1080p YUYV` 并选择该分辨率申报的最高帧率；如果没有 1080p YUYV，立即将 `video_input_select` 标记为 `FAIL` 并中止测试，不再回退到较低分辨率。
 - CSI/MIPI：只测试一套 `1080p60 NV12`，不做输入格式切换。
 
 每个输入配置都会跑三种输出：
@@ -87,6 +87,7 @@ python okvm_testctl.py run `
 - H.265 WebRTC
 
 默认每个视频输出模式采样 30 秒；可通过 `--sample-seconds <秒数>` 覆盖。
+每次应用视频输入配置后默认先空转 3 秒，稳定后再开始统计；可通过 `--video-config-settle-seconds <秒数>` 调整。
 
 MJPEG/HTTP 测试时，控制端会让 Windows agent 输出默认 60fps 的全屏动态画面，避免静态画面触发 MJPEG “无变化不发帧”策略导致 fps 误判；可通过 `--mjpeg-motion-fps <fps>` 覆盖。
 

test/okvm-test/agent/main.go

@@ -42,7 +42,6 @@ const (
 	wmChar        = 0x0102
 	wmSysKeyDown  = 0x0104
 	wmSysKeyUp    = 0x0105
-	wmTimer       = 0x0113
 	wmMouseMove   = 0x0200
 	wmLButtonDown = 0x0201
 	wmLButtonUp   = 0x0202
@@ -75,10 +74,11 @@ const (
 	wmAppFocus        = wmApp + 4
 	wmAppDynamicStart = wmApp + 5
 	wmAppDynamicStop  = wmApp + 6
-
+	wmAppDynamicFrame = wmApp + 7
-	dynamicTimerID = 1
 
 	colorWindow            = 5
+	dynamicBackgroundColor = 0x00101010
+	dynamicPatchSize       = 384
 
 	driveUnknown   = 0
 	driveNoRootDir = 1
@@ -111,16 +111,17 @@ var (
 	procShowWindow            = user32.NewProc("ShowWindow")
 	procSetForegroundWindow   = user32.NewProc("SetForegroundWindow")
 	procGetSystemMetrics      = user32.NewProc("GetSystemMetrics")
+	procGetDC                 = user32.NewProc("GetDC")
+	procReleaseDC             = user32.NewProc("ReleaseDC")
 	procInvalidateRect        = user32.NewProc("InvalidateRect")
 	procUpdateWindow          = user32.NewProc("UpdateWindow")
-	procSetTimer              = user32.NewProc("SetTimer")
-	procKillTimer             = user32.NewProc("KillTimer")
 	procGetKeyState           = user32.NewProc("GetKeyState")
 	procBeginPaint            = user32.NewProc("BeginPaint")
 	procEndPaint              = user32.NewProc("EndPaint")
 	procFillRect              = user32.NewProc("FillRect")
 	procCreateSolidBrush      = gdi32.NewProc("CreateSolidBrush")
 	procDeleteObject          = gdi32.NewProc("DeleteObject")
+	procGetDeviceCaps         = gdi32.NewProc("GetDeviceCaps")
 	procImmAssociateContext   = imm32.NewProc("ImmAssociateContext")
 	procGetModuleHandleW      = kernel32.NewProc("GetModuleHandleW")
 	procQueryPerformanceCount = kernel32.NewProc("QueryPerformanceCounter")
@@ -192,6 +193,10 @@ type appState struct {
 	dynamicActive       bool
 	dynamicFPS          int
 	dynamicFrame        int64
+	dynamicGeneration   uint64
+	dynamicFramePending bool
+	dynamicStop         chan struct{}
+	dynamicStarted      time.Time
 	events              []hidEvent
 }
 
@@ -519,20 +524,22 @@ func windowProc(hwnd uintptr, message uintptr, wParam, lParam uintptr) uintptr {
 		hdc, _, _ := procBeginPaint.Call(hwnd, uintptr(unsafe.Pointer(&ps)))
 		state.mu.Lock()
 		color := state.bgColor
+		dynamic := state.dynamicActive
 		state.mu.Unlock()
-		brush, _, _ := procCreateSolidBrush.Call(uintptr(color))
+		paintRect := ps.RcPaint
-		r := rect{Left: 0, Top: 0, Right: int32(screenWidth()), Bottom: int32(screenHeight())}
+		if paintRect.Right <= paintRect.Left || paintRect.Bottom <= paintRect.Top {
-		procFillRect.Call(hdc, uintptr(unsafe.Pointer(&r)), brush)
+			paintRect = rect{Left: 0, Top: 0, Right: int32(screenWidth()), Bottom: int32(screenHeight())}
-		procDeleteObject.Call(brush)
+		}
+		if dynamic {
+			fillRect(hdc, paintRect, dynamicBackgroundColor)
+			if patch, ok := intersectRects(paintRect, dynamicPatchRect()); ok {
+				fillRect(hdc, patch, color)
+			}
+		} else {
+			fillRect(hdc, paintRect, color)
+		}
 		procEndPaint.Call(hwnd, uintptr(unsafe.Pointer(&ps)))
 		return 0
-	case wmTimer:
-		if wParam == dynamicTimerID {
-			advanceDynamicFrame(hwnd)
-			return 0
-		}
-		ret, _, _ := procDefWindowProcW.Call(hwnd, message, wParam, lParam)
-		return ret
 	case wmAppInvalidate:
 		invalidateWindowNow(hwnd)
 		return 0
@@ -547,16 +554,12 @@ func windowProc(hwnd uintptr, message uintptr, wParam, lParam uintptr) uintptr {
 		procSetForegroundWindow.Call(hwnd)
 		return 0
 	case wmAppDynamicStart:
-		procKillTimer.Call(hwnd, dynamicTimerID)
-		interval := wParam
-		if interval == 0 {
-			interval = 16
-		}
-		procSetTimer.Call(hwnd, dynamicTimerID, interval, 0)
 		invalidateWindowNow(hwnd)
 		return 0
+	case wmAppDynamicFrame:
+		advanceDynamicFrame(hwnd, uint64(wParam))
+		return 0
 	case wmAppDynamicStop:
-		procKillTimer.Call(hwnd, dynamicTimerID)
 		invalidateWindowNow(hwnd)
 		return 0
 	case wmKeyDown:
@@ -590,7 +593,6 @@ func windowProc(hwnd uintptr, message uintptr, wParam, lParam uintptr) uintptr {
 		procShowWindow.Call(hwnd, swHide)
 		return 0
 	case wmDestroy:
-		procKillTimer.Call(hwnd, dynamicTimerID)
 		procPostQuitMessage.Call(0)
 		return 0
 	default:
@@ -704,13 +706,20 @@ func startDynamic(fps int) map[string]interface{} {
 		fps = 120
 	}
 	stopDynamic()
+	stop := make(chan struct{})
 	state.mu.Lock()
 	state.dynamicActive = true
 	state.dynamicFPS = fps
 	state.dynamicFrame = 0
+	state.dynamicGeneration++
+	generation := state.dynamicGeneration
+	state.dynamicFramePending = false
+	state.dynamicStop = stop
+	state.dynamicStarted = time.Now()
 	display := setColorStateLocked(dynamicFrameColor(0))
 	state.mu.Unlock()
-	postUIMessage(wmAppDynamicStart, uintptr(dynamicTimerIntervalMS(fps)), 0)
+	postUIMessage(wmAppDynamicStart, uintptr(generation), 0)
+	go runDynamicFrames(fps, generation, stop)
 	display["dynamic"] = true
 	display["fps"] = fps
 	return display
@@ -719,24 +728,51 @@ func startDynamic(fps int) map[string]interface{} {
 func stopDynamic() {
 	state.mu.Lock()
 	active := state.dynamicActive
+	stop := state.dynamicStop
 	state.dynamicActive = false
 	state.dynamicFPS = 0
 	state.dynamicFrame = 0
+	state.dynamicFramePending = false
+	state.dynamicStop = nil
+	state.dynamicStarted = time.Time{}
 	state.mu.Unlock()
+	if stop != nil {
+		close(stop)
+	}
 	if active {
 		postUIMessage(wmAppDynamicStop, 0, 0)
 	}
 }
 
-func dynamicTimerIntervalMS(fps int) int {
+func runDynamicFrames(fps int, generation uint64, stop <-chan struct{}) {
-	if fps < 1 {
+	interval := time.Second / time.Duration(fps)
-		fps = 60
+	ticker := time.NewTicker(interval)
+	defer ticker.Stop()
+	for {
+		select {
+		case <-stop:
+			return
+		case <-ticker.C:
+			state.mu.Lock()
+			if !state.dynamicActive || state.dynamicGeneration != generation {
+				state.mu.Unlock()
+				return
+			}
+			if state.dynamicFramePending {
+				state.mu.Unlock()
+				continue
+			}
+			state.dynamicFramePending = true
+			state.mu.Unlock()
+			if !postUIMessage(wmAppDynamicFrame, uintptr(generation), 0) {
+				state.mu.Lock()
+				if state.dynamicGeneration == generation {
+					state.dynamicFramePending = false
+				}
+				state.mu.Unlock()
+			}
 		}
-	interval := 1000 / fps
-	if interval < 1 {
-		return 1
 	}
-	return interval
 }
 
 func dynamicFrameColor(frame int64) string {
@@ -746,16 +782,17 @@ func dynamicFrameColor(frame int64) string {
 	return fmt.Sprintf("#%02x%02x%02x", r, g, b)
 }
 
-func advanceDynamicFrame(hwnd uintptr) {
+func advanceDynamicFrame(hwnd uintptr, generation uint64) {
 	state.mu.Lock()
-	if !state.dynamicActive {
+	if !state.dynamicActive || state.dynamicGeneration != generation {
 		state.mu.Unlock()
 		return
 	}
+	state.dynamicFramePending = false
 	state.dynamicFrame++
 	setColorStateLocked(dynamicFrameColor(state.dynamicFrame))
 	state.mu.Unlock()
-	invalidateWindowNow(hwnd)
+	invalidateDynamicPatchNow(hwnd)
 }
 
 func setColorStateLocked(colorHex string) map[string]interface{} {
@@ -781,6 +818,13 @@ func setColorStateLocked(colorHex string) map[string]interface{} {
 func currentDisplayState() map[string]interface{} {
 	state.mu.Lock()
 	defer state.mu.Unlock()
+	dynamicActualFPS := 0.0
+	if state.dynamicActive && !state.dynamicStarted.IsZero() {
+		elapsed := time.Since(state.dynamicStarted).Seconds()
+		if elapsed > 0 {
+			dynamicActualFPS = float64(state.dynamicFrame) / elapsed
+		}
+	}
 	return map[string]interface{}{
 		"color":                 state.colorHex,
 		"last_change_qpc":       state.lastColorChangeQpc,
@@ -788,6 +832,9 @@ func currentDisplayState() map[string]interface{} {
 		"sequence":              state.colorSequence,
 		"dynamic":               state.dynamicActive,
 		"dynamic_fps":           state.dynamicFPS,
+		"dynamic_frame":         state.dynamicFrame,
+		"dynamic_actual_fps":    dynamicActualFPS,
+		"display_refresh_hz":    screenRefreshHz(),
 		"qpc":                   qpcNow(),
 		"unix_nano":             time.Now().UnixNano(),
 	}
@@ -816,13 +863,15 @@ func focusWindow() {
 	}
 }
 
-func postUIMessage(message uint32, wParam uintptr, lParam uintptr) {
+func postUIMessage(message uint32, wParam uintptr, lParam uintptr) bool {
 	state.mu.Lock()
 	hwnd := state.hwnd
 	state.mu.Unlock()
-	if hwnd != 0 {
+	if hwnd == 0 {
-		procPostMessageW.Call(hwnd, uintptr(message), wParam, lParam)
+		return false
 	}
+	result, _, _ := procPostMessageW.Call(hwnd, uintptr(message), wParam, lParam)
+	return result != 0
 }
 
 func invalidateWindowNow(hwnd uintptr) {
@@ -830,6 +879,60 @@ func invalidateWindowNow(hwnd uintptr) {
 	procUpdateWindow.Call(hwnd)
 }
 
+func invalidateDynamicPatchNow(hwnd uintptr) {
+	patch := dynamicPatchRect()
+	procInvalidateRect.Call(hwnd, uintptr(unsafe.Pointer(&patch)), 0)
+	procUpdateWindow.Call(hwnd)
+}
+
+func dynamicPatchRect() rect {
+	width := int32(screenWidth())
+	height := int32(screenHeight())
+	size := int32(dynamicPatchSize)
+	if size > width {
+		size = width
+	}
+	if size > height {
+		size = height
+	}
+	left := (width - size) / 2
+	top := (height - size) / 2
+	return rect{Left: left, Top: top, Right: left + size, Bottom: top + size}
+}
+
+func intersectRects(a, b rect) (rect, bool) {
+	intersection := rect{
+		Left:   maxInt32(a.Left, b.Left),
+		Top:    maxInt32(a.Top, b.Top),
+		Right:  minInt32(a.Right, b.Right),
+		Bottom: minInt32(a.Bottom, b.Bottom),
+	}
+	return intersection, intersection.Right > intersection.Left && intersection.Bottom > intersection.Top
+}
+
+func fillRect(hdc uintptr, area rect, color uint32) {
+	brush, _, _ := procCreateSolidBrush.Call(uintptr(color))
+	if brush == 0 {
+		return
+	}
+	procFillRect.Call(hdc, uintptr(unsafe.Pointer(&area)), brush)
+	procDeleteObject.Call(brush)
+}
+
+func minInt32(a, b int32) int32 {
+	if a < b {
+		return a
+	}
+	return b
+}
+
+func maxInt32(a, b int32) int32 {
+	if a > b {
+		return a
+	}
+	return b
+}
+
 func disableIME(hwnd uintptr) {
 	if hwnd != 0 {
 		procImmAssociateContext.Call(hwnd, 0)
@@ -870,7 +973,18 @@ func screenSize() (int, int) {
 
 func screenInfo() map[string]int {
 	width, height := screenSize()
-	return map[string]int{"width": width, "height": height}
+	return map[string]int{"width": width, "height": height, "refresh_hz": screenRefreshHz()}
+}
+
+func screenRefreshHz() int {
+	const vRefresh = 116
+	hdc, _, _ := procGetDC.Call(0)
+	if hdc == 0 {
+		return 0
+	}
+	defer procReleaseDC.Call(0, hdc)
+	refresh, _, _ := procGetDeviceCaps.Call(hdc, vRefresh)
+	return int(refresh)
 }
 
 func screenWidth() int {

test/okvm-test/okvm_report.py

@@ -165,7 +165,7 @@ class Reporter:
             "- 测试设备：",
             f"- 视频设备：{markdown_inline(self.user_video_device())}",
             f"- HID设备：{markdown_inline(self.user_hid_backend())}",
-            f"- 网络延迟：{markdown_inline(self.user_network_latency())}",
+            f"- HTTP 延迟：{markdown_inline(self.user_http_latency())}",
             "",
             "## 视频性能",
             "",
@@ -229,14 +229,14 @@ class Reporter:
                 return result
         return None
 
-    def user_network_latency(self) -> str:
+    def user_http_latency(self) -> str:
         result = self.find_result("network_latency")
         if not result:
             return "无数据"
-        tcp = result.data.get("tcp_connect") or {}
+        http = result.data.get("http_health") or {}
-        if tcp.get("samples"):
+        if http.get("samples"):
-            return format_latency_values(tcp)
+            return format_latency_values(http)
-        return result.message or "无数据"
+        return "无数据"
 
     def user_video_device(self) -> str:
         result = self.find_result("video_input_select")

test/okvm-test/okvm_testctl.py

@@ -275,6 +275,24 @@ class DeviceSelector:
                     return width, height, best
         return None
 
+    @staticmethod
+    def _pick_required_mode(
+        fmt: dict[str, Any] | None,
+        width: int,
+        height: int,
+        fps: float,
+    ) -> tuple[int, int, float] | None:
+        if not fmt:
+            return None
+        for res in fmt.get("resolutions", []):
+            if int(res.get("width", 0)) != width or int(res.get("height", 0)) != height:
+                continue
+            for advertised_fps in res.get("fps", []):
+                value = float(advertised_fps)
+                if abs(value - fps) <= 0.05:
+                    return width, height, value
+        return None
+
     @staticmethod
     def _pick_highest_1080(fmt: dict[str, Any]) -> tuple[int, int, float] | None:
         candidates: list[tuple[int, int, float]] = []
@@ -289,6 +307,22 @@ class DeviceSelector:
             return None
         return max(candidates, key=lambda x: (x[0] * x[1], x[2]))
 
+    @staticmethod
+    def _pick_highest_fps_at_resolution(
+        fmt: dict[str, Any] | None,
+        width: int,
+        height: int,
+    ) -> tuple[int, int, float] | None:
+        if not fmt:
+            return None
+        candidates: list[float] = []
+        for res in fmt.get("resolutions", []):
+            if int(res.get("width", 0)) == width and int(res.get("height", 0)) == height:
+                candidates.extend(float(value) for value in res.get("fps", []))
+        if not candidates:
+            return None
+        return width, height, max(candidates)
+
     def select(self) -> list[VideoInputCase]:
         video_devices = self.devices.get("video", [])
         if not video_devices:
@@ -310,17 +344,36 @@ class DeviceSelector:
 
         mjpeg = self._find_format(device, "MJPEG")
         yuyv = self._find_format(device, "YUYV")
-        target_fps = 60 if input_class == "usb3" else 30
+        if input_class == "usb2":
+            required_modes = (
+                ("MJPEG", mjpeg, 50.0, "usb2_mjpeg"),
+                ("YUYV", yuyv, 10.0, "usb2_yuyv"),
+            )
+            missing: list[str] = []
+            for fmt_name, fmt_info, required_fps, label in required_modes:
+                picked = self._pick_required_mode(fmt_info, 1920, 1080, required_fps)
+                if not picked:
+                    missing.append(f"{fmt_name} 1920x1080@{required_fps:g}fps")
+                    continue
+                width, height, fps = picked
+                cases.append(VideoInputCase(label, input_class, path, fmt_name, width, height, fps))
+            if missing:
+                raise RuntimeError(
+                    "USB 2.0 capture card is missing required mode(s): " + ", ".join(missing)
+                )
+            return cases
+
+        target_fps = 60
         if mjpeg:
             picked = self._pick_exact(mjpeg, 1920, 1080, target_fps) or self._pick_highest_1080(mjpeg)
             if picked:
                 w, h, f = picked
                 cases.append(VideoInputCase(f"{input_class}_mjpeg", input_class, path, "MJPEG", w, h, f))
-        if yuyv:
+        picked = self._pick_highest_fps_at_resolution(yuyv, 1920, 1080)
-            picked = self._pick_highest_1080(yuyv)
+        if not picked:
-            if picked:
+            raise RuntimeError("USB 3.0 capture card is missing required mode: YUYV 1920x1080")
         w, h, f = picked
-                cases.append(VideoInputCase(f"{input_class}_yuyv", input_class, path, "YUYV", w, h, f))
+        cases.append(VideoInputCase("usb3_yuyv", input_class, path, "YUYV", w, h, f))
         return cases
 
 
@@ -597,7 +650,16 @@ echo "$BACKUP"
 
     def select_video_cases(self, devices: dict[str, Any]) -> list[VideoInputCase]:
         selector = DeviceSelector(self.lsusb_tree, devices)
+        try:
             cases = selector.select()
+        except RuntimeError as exc:
+            self.reporter.add(
+                "video_input_select",
+                "FAIL",
+                str(exc),
+                devices=devices.get("video", []),
+            )
+            raise
         if not cases:
             self.reporter.add("video_input_select", "FAIL", "no suitable video input case found", devices=devices.get("video", []))
             return []
@@ -692,7 +754,6 @@ echo "$BACKUP"
                     )
                     continue
                 try:
-                    self.apply_video_case(case)
                     if output_mode == "mjpeg":
                         motion_started = await self.start_mjpeg_motion()
                         try:
@@ -796,6 +857,9 @@ echo "$BACKUP"
                 "quality": self.args.jpeg_quality,
             },
         )
+        settle_seconds = max(0.0, float(self.args.video_config_settle_seconds))
+        if settle_seconds > 0:
+            time.sleep(settle_seconds)
 
     def configure_video_case(self, case: VideoInputCase) -> None:
         self.apply_video_case(case)
@@ -818,7 +882,7 @@ echo "$BACKUP"
         frame_count = 0
         byte_count = 0
         first_frame_s: float | None = None
-        for frame, frame_time, _ in self.iter_mjpeg_frames(client_id, timeout=self.args.sample_seconds, video_case=case):
+        for frame, frame_time, _ in self.iter_mjpeg_frames(client_id, timeout=self.args.sample_seconds):
             now = time.monotonic()
             if now >= deadline:
                 break
@@ -855,6 +919,7 @@ echo "$BACKUP"
 
     async def measure_webrtc(self, case: VideoInputCase, codec: str) -> dict[str, Any]:
         self.set_stream_mode(codec)
+        self.apply_video_case(case)
         try:
             from playwright.async_api import async_playwright
         except ImportError:
@@ -885,7 +950,13 @@ echo "$BACKUP"
             try:
                 page = await context.new_page()
                 await page.goto(self.api.base)
-                result = await page.evaluate(js, {"seconds": self.args.sample_seconds})
+                result = await page.evaluate(
+                    js,
+                    {
+                        "seconds": self.args.sample_seconds,
+                        "settleSeconds": max(0.0, float(self.args.video_config_settle_seconds)),
+                    },
+                )
             except Exception as exc:
                 text = str(exc)
                 if codec == "h265" and is_webrtc_codec_unsupported_error(text):
@@ -988,7 +1059,6 @@ echo "$BACKUP"
                     )
                     continue
                 try:
-                    self.apply_video_case(latency_case)
                     if output_mode == "mjpeg":
                         await self.run_mjpeg_latency_test(latency_case, check_name=check_name, save_evidence=False)
                     else:
@@ -1010,17 +1080,8 @@ echo "$BACKUP"
         return cases[0]
 
     def configure_hdmi_probe_case(self, case: VideoInputCase) -> None:
-        self.api.patch(
+        self.set_stream_mode("mjpeg")
-            "/config/video",
+        self.apply_video_case(case)
-            {
-                "device": case.device,
-                "format": case.fmt,
-                "width": case.width,
-                "height": case.height,
-                "fps": int(round(case.fps)),
-                "quality": self.args.jpeg_quality,
-            },
-        )
         self.reporter.add(
             "config_video_hdmi_probe",
             "PASS",
@@ -1044,7 +1105,6 @@ echo "$BACKUP"
                 evidence_title=f"HDMI 纯色采集帧 {name}",
                 expected_rgb=expected,
                 threshold=self.args.hdmi_color_fail,
-                video_case=case,
             )
             err = rgb_error(expected, stats["mean_rgb"])
             closest_name, closest_error = closest_hdmi_color(stats["mean_rgb"])
@@ -1113,6 +1173,9 @@ echo "$BACKUP"
         if self.args.hdmi_latency_trials <= 0:
             self.reporter.add(check_name, "SKIP", "video latency trials disabled", video_case=case.__dict__, output_mode="mjpeg")
             return
+        self.set_stream_mode("mjpeg")
+        self.apply_video_case(case)
+        self.api.post("/stream/start", {})
         offset_ns, sync = await self.sync_agent_clock(f"{check_name}_agent_clock_sync_rtt")
         trials: list[dict[str, Any]] = []
         colors = [("#ff0000", "#00ff00"), ("#00ff00", "#0000ff"), ("#0000ff", "#ff0000")]
@@ -1128,7 +1191,6 @@ echo "$BACKUP"
                 timeout=detect_timeout,
                 threshold=self.args.hdmi_color_fail,
                 evidence_title=f"HDMI 延迟命中帧 {case.label} #{i + 1}" if save_evidence else None,
-                video_case=case,
             )
             display = await self.agent.command("display_state", {}, timeout=5)
             actual_agent_ns = int(display.get("last_change_unix_nano") or 0)
@@ -1209,6 +1271,9 @@ echo "$BACKUP"
                 setup = await page.evaluate(WEBRTC_LATENCY_SETUP_JS, {"timeoutMs": 15000})
                 if not setup.get("connected"):
                     raise RuntimeError(f"{output_mode} WebRTC did not connect: {setup}")
+                settle_seconds = max(0.0, float(self.args.video_config_settle_seconds))
+                if settle_seconds > 0:
+                    await page.wait_for_timeout(settle_seconds * 1000)
                 for i in range(self.args.hdmi_latency_trials):
                     source, target = colors[i % len(colors)]
                     await self.agent.command("show", {"color": source, "full": True}, timeout=5)
@@ -1313,7 +1378,6 @@ echo "$BACKUP"
         evidence_title: str | None = None,
         expected_rgb: tuple[int, int, int] | None = None,
         threshold: float | None = None,
-        video_case: VideoInputCase | None = None,
     ) -> dict[str, Any]:
         threshold = self.args.hdmi_color_fail if threshold is None else threshold
         required_matches = max(1, int(self.args.hdmi_match_frames))
@@ -1324,7 +1388,7 @@ echo "$BACKUP"
         consecutive_matches = 0
         frames_seen = 0
 
-        for frame, _, wall_ns in self.iter_mjpeg_frames(client_label, timeout, video_case=video_case):
+        for frame, _, wall_ns in self.iter_mjpeg_frames(client_label, timeout):
             frames_seen += 1
             stats = jpeg_rgb_stats(frame)
             stats["wall_ns"] = wall_ns
@@ -1370,12 +1434,11 @@ echo "$BACKUP"
         timeout: float,
         threshold: float,
         evidence_title: str | None = None,
-        video_case: VideoInputCase | None = None,
     ) -> dict[str, Any]:
         last: dict[str, Any] | None = None
         last_frame: bytes | None = None
         frames_seen = 0
-        for frame, _, wall_ns in self.iter_mjpeg_frames(client_label, timeout, video_case=video_case):
+        for frame, _, wall_ns in self.iter_mjpeg_frames(client_label, timeout):
             frames_seen += 1
             stats = jpeg_rgb_stats(frame)
             err = rgb_error(expected_rgb, stats["mean_rgb"])
@@ -1400,10 +1463,8 @@ echo "$BACKUP"
         path.write_bytes(frame)
         self.reporter.add_evidence("HDMI 采集帧", title, path)
 
-    def iter_mjpeg_frames(self, client_label: str, timeout: float, video_case: VideoInputCase | None = None):
+    def iter_mjpeg_frames(self, client_label: str, timeout: float):
         self.set_stream_mode("mjpeg")
-        if video_case is not None:
-            self.apply_video_case(video_case)
         self.api.post("/stream/start", {})
         client_id = f"test-{self.run_id}-{client_label}"
         url = f"{self.api.base}/api/stream/mjpeg?client_id={client_id}"
@@ -1441,8 +1502,6 @@ echo "$BACKUP"
                     raise
                 time.sleep(0.5)
                 self.set_stream_mode("mjpeg", timeout=10)
-                if video_case is not None:
-                    self.apply_video_case(video_case)
                 self.api.post("/stream/start", {})
 
     async def run_hid_test(self) -> None:
@@ -1834,7 +1893,7 @@ VIDEO_SCREENSHOT_READY_JS = r"""
 
 
 WEBRTC_MEASURE_JS = r"""
-async ({seconds}) => {
+async ({seconds, settleSeconds}) => {
   const api = async (path, opts = {}) => {
     const response = await fetch('/api' + path, {
       credentials: 'include',
@@ -1895,6 +1954,12 @@ async ({seconds}) => {
     await new Promise(r => setTimeout(r, 100));
   }
 
+  // Let capture, encoding, decoding, and browser rendering reach steady state.
+  // Samples collected during this interval are intentionally discarded.
+  if (settleSeconds > 0) {
+    await new Promise(r => setTimeout(r, settleSeconds * 1000));
+  }
+
   const samples = [];
   const endAt = performance.now() + seconds * 1000;
   while (performance.now() < endAt) {
@@ -2252,6 +2317,7 @@ def build_parser() -> argparse.ArgumentParser:
     run.add_argument("--network-latency-samples", type=int, default=7, help="controller-to-target network latency samples collected during setup")
     run.add_argument("--network-latency-timeout", type=float, default=3.0, help="per-sample TCP/HTTP latency timeout in seconds")
     run.add_argument("--sample-seconds", type=int, default=30)
+    run.add_argument("--video-config-settle-seconds", type=float, default=3.0, help="idle time after applying video configuration before collecting samples")
     run.add_argument("--jpeg-quality", type=int, default=80)
     run.add_argument("--mjpeg-motion-fps", type=int, default=60, help="Windows agent dynamic source fps used during MJPEG/HTTP tests")
     run.add_argument("--agent-host", default=None, help="Windows agent IP/hostname; omit to skip Windows-side checks")