win-dshow: Support bottom-up DIBs

The biHeight field can be negative, leading to crashes on some cards
like VisionRGB-E1S. Adding flip support is fairly straightforward.

There also appears to be a hack to automatically flip for RGB formats,
but I wish to remove it because it seems to fight with this change. We
already have a separate vertical flip checkbox to deal with non
compliant behavior.

plugins/win-dshow/libdshowcapture

@@ -1 +1 @@
-Subproject commit 13cff539a4b4dfbda3dbf98b4f9863dc5295e603
+Subproject commit d23cce714d38483053bdc6e191df029a32896cd3

plugins/win-dshow/win-dshow.cpp

@@ -479,7 +479,8 @@ void DShowInput::OnEncodedVideoData(enum AVCodecID id, unsigned char *data,
 		 * than 1920x1080.  The reason why is because we want to strike
 		 * a reasonable balance between hardware and CPU usage. */
 		bool useHW = videoConfig.format != VideoFormat::MJPEG ||
-			     (videoConfig.cx * videoConfig.cy) > MAX_SW_RES_INT;
+			     (videoConfig.cx * videoConfig.cy_abs) >
+				     MAX_SW_RES_INT;
 		if (ffmpeg_decode_init(video_decoder, id, useHW) < 0) {
 			blog(LOG_WARNING, "Could not initialize video decoder");
 			return;
@@ -520,16 +521,15 @@ void DShowInput::OnVideoData(const VideoConfig &config, unsigned char *data,
 	}
 
 	const int cx = config.cx;
-	const int cy = config.cy;
+	const int cy_abs = config.cy_abs;
 
 	frame.timestamp = (uint64_t)startTime * 100;
 	frame.width = config.cx;
-	frame.height = config.cy;
+	frame.height = cy_abs;
 	frame.format = ConvertVideoFormat(config.format);
-	frame.flip = (config.format == VideoFormat::XRGB ||
-		      config.format == VideoFormat::ARGB);
+	frame.flip = flip;
 
-	if (flip)
+	if (config.cy_flip)
 		frame.flip = !frame.flip;
 
 	if (videoConfig.format == VideoFormat::XRGB ||
@@ -546,23 +546,23 @@ void DShowInput::OnVideoData(const VideoConfig &config, unsigned char *data,
 
 	} else if (videoConfig.format == VideoFormat::I420) {
 		frame.data[0] = data;
-		frame.data[1] = frame.data[0] + (cx * cy);
-		frame.data[2] = frame.data[1] + (cx * cy / 4);
+		frame.data[1] = frame.data[0] + (cx * cy_abs);
+		frame.data[2] = frame.data[1] + (cx * cy_abs / 4);
 		frame.linesize[0] = cx;
 		frame.linesize[1] = cx / 2;
 		frame.linesize[2] = cx / 2;
 
 	} else if (videoConfig.format == VideoFormat::YV12) {
 		frame.data[0] = data;
-		frame.data[2] = frame.data[0] + (cx * cy);
-		frame.data[1] = frame.data[2] + (cx * cy / 4);
+		frame.data[2] = frame.data[0] + (cx * cy_abs);
+		frame.data[1] = frame.data[2] + (cx * cy_abs / 4);
 		frame.linesize[0] = cx;
 		frame.linesize[1] = cx / 2;
 		frame.linesize[2] = cx / 2;
 
 	} else if (videoConfig.format == VideoFormat::NV12) {
 		frame.data[0] = data;
-		frame.data[1] = frame.data[0] + (cx * cy);
+		frame.data[1] = frame.data[0] + (cx * cy_abs);
 		frame.linesize[0] = cx;
 		frame.linesize[1] = cx;
 
@@ -910,7 +910,8 @@ bool DShowInput::UpdateVideoConfig(obs_data_t *settings)
 	videoConfig.path = id.path.c_str();
 	videoConfig.useDefaultConfig = resType == ResType_Preferred;
 	videoConfig.cx = cx;
-	videoConfig.cy = cy;
+	videoConfig.cy_abs = abs(cy);
+	videoConfig.cy_flip = cy < 0;
 	videoConfig.frameInterval = interval;
 	videoConfig.internalFormat = format;
 
@@ -950,11 +951,13 @@ bool DShowInput::UpdateVideoConfig(obs_data_t *settings)
 	     "\tvideo device: %s\n"
 	     "\tvideo path: %s\n"
 	     "\tresolution: %dx%d\n"
+	     "\tflip: %d\n"
 	     "\tfps: %0.2f (interval: %lld)\n"
 	     "\tformat: %s",
 	     obs_source_get_name(source), (const char *)name_utf8,
-	     (const char *)path_utf8, videoConfig.cx, videoConfig.cy, fps,
-	     videoConfig.frameInterval, formatName->array);
+	     (const char *)path_utf8, videoConfig.cx, videoConfig.cy_abs,
+	     (int)videoConfig.cy_flip, fps, videoConfig.frameInterval,
+	     formatName->array);
 
 	SetupBuffering(settings);