Remove video framerate as explicit constraint (#394)

* Do not consider video framerate in the constraints fitness function
* Remove test about video framerate

mediadevices_test.go

@@ -125,18 +125,6 @@ func TestSelectBestDriverConstraintsResultIsSetProperly(t *testing.T) {
 			frameFormat: frame.FormatI420,
 			frameRate:   expectedProp.FrameRate,
 		},
-		"DifferentFrameRate": {
-			width:       expectedProp.Width,
-			height:      expectedProp.Height,
-			frameFormat: expectedProp.FrameFormat,
-			frameRate:   expectedProp.FrameRate - 1,
-		},
-		"NoFrameRateConstraints": {
-			width:       expectedProp.Width,
-			height:      expectedProp.Height,
-			frameFormat: expectedProp.FrameFormat,
-			frameRate:   -1,
-		},
 	}
 
 	for name, c := range cases {
@@ -229,12 +217,6 @@ func TestSelectBestDriverConstraintsNoFit(t *testing.T) {
 			frameFormat: frame.FormatI420,
 			frameRate:   expectedProp.FrameRate,
 		},
-		"DifferentFrameRate": {
-			width:       expectedProp.Width,
-			height:      expectedProp.Height,
-			frameFormat: expectedProp.FrameFormat,
-			frameRate:   expectedProp.FrameRate - 1,
-		},
 	}
 
 	for name, c := range cases {

pkg/codec/openh264/bridge.cpp

@@ -81,7 +81,7 @@ Slice enc_encode(Encoder *e, Frame f, int *eresult) {
   Slice payload = {0};
 
   if(e->force_key_frame == 1) {
-    info.eFrameType = videoFrameTypeI;
+    e->engine->ForceIntraFrame(true);
     e->force_key_frame = 0;
   }
 

pkg/codec/x264/bridge.h

@@ -18,6 +18,7 @@ typedef struct Encoder {
   x264_t *h;
   x264_picture_t pic_in;
   x264_param_t param;
+  int force_key_frame;
 } Encoder;
 
 Encoder *enc_new(x264_param_t param, char *preset, int *rc) {
@@ -85,8 +86,14 @@ Slice enc_encode(Encoder *e, uint8_t *y, uint8_t *cb, uint8_t *cr, int *rc) {
   e->pic_in.img.plane[0] = y;
   e->pic_in.img.plane[1] = cb;
   e->pic_in.img.plane[2] = cr;
+  if (e->force_key_frame) {
+    e->pic_in.i_type = X264_TYPE_IDR;
+  } else {
+    e->pic_in.i_type = X264_TYPE_AUTO;
+  }
 
   int frame_size = x264_encoder_encode(e->h, &nal, &i_nal, &e->pic_in, &pic_out);
+  e->force_key_frame = 0;
   Slice s = {.data_len = frame_size};
   if (frame_size <= 0) {
     *rc = ERR_ENCODE;

pkg/codec/x264/x264.go

@@ -129,7 +129,8 @@ func (e *encoder) SetBitRate(b int) error {
 }
 
 func (e *encoder) ForceKeyFrame() error {
-	panic("ForceKeyFrame is not implemented")
+	e.engine.force_key_frame = C.int(1)
+	return nil
 }
 
 func (e *encoder) Close() error {

pkg/driver/vncdriver/vnc/encoding.go

@@ -18,13 +18,32 @@ type Encoding interface {
 	// the proper data.
 	Read(*ClientConn, *Rectangle, io.Reader) (Encoding, error)
 }
+type CursorEncoding struct {
+}
+
+func (*CursorEncoding) Type() int32 {
+	return -239
+}
+func (*CursorEncoding) Read(c *ClientConn, rect *Rectangle, r io.Reader) (Encoding, error) {
+	size := int(rect.Height) * int(rect.Width) * int(c.PixelFormat.BPP) / 8
+	pixelBytes := make([]uint8, size)
+	if _, err := io.ReadFull(r, pixelBytes); err != nil {
+		return nil, err
+	}
+	mask := ((int(rect.Width) + 7) / 8) * int(rect.Height)
+	maskBytes := make([]uint8, mask)
+	if _, err := io.ReadFull(r, maskBytes); err != nil {
+		return nil, err
+	}
+	return &CursorEncoding{}, nil
+}
 
 // RawEncoding is raw pixel data sent by the server.
 //
 // See RFC 6143 Section 7.7.1
 type RawEncoding struct {
-	Colors []Color
-	RawPixel []uint32	//RGBA
+	Colors   []Color
+	RawPixel []uint32 //RGBA
 }
 
 func (*RawEncoding) Type() int32 {
@@ -32,6 +51,7 @@ func (*RawEncoding) Type() int32 {
 }
 
 func (*RawEncoding) Read(c *ClientConn, rect *Rectangle, r io.Reader) (Encoding, error) {
+	//fmt.Println("RawEncoding")
 	bytesPerPixel := c.PixelFormat.BPP / 8
 	pixelBytes := make([]uint8, bytesPerPixel)
 
@@ -41,7 +61,7 @@ func (*RawEncoding) Read(c *ClientConn, rect *Rectangle, r io.Reader) (Encoding,
 	}
 
 	colors := make([]Color, int(rect.Height)*int(rect.Width))
-	rawPixels:=make([]uint32,int(rect.Height)*int(rect.Width))
+	rawPixels := make([]uint32, int(rect.Height)*int(rect.Width))
 	for y := uint16(0); y < rect.Height; y++ {
 		for x := uint16(0); x < rect.Width; x++ {
 			if _, err := io.ReadFull(r, pixelBytes); err != nil {
@@ -62,24 +82,30 @@ func (*RawEncoding) Read(c *ClientConn, rect *Rectangle, r io.Reader) (Encoding,
 				color.R = uint16((rawPixel >> c.PixelFormat.RedShift) & uint32(c.PixelFormat.RedMax))
 				color.G = uint16((rawPixel >> c.PixelFormat.GreenShift) & uint32(c.PixelFormat.GreenMax))
 				color.B = uint16((rawPixel >> c.PixelFormat.BlueShift) & uint32(c.PixelFormat.BlueMax))
+				if c.PixelFormat.BPP == 16 {
+					color.B = color.B<<3 | color.B>>2
+					color.G = color.G<<2 | color.G>>2
+					color.R = color.R<<3 | color.R>>2
+				}
 			} else {
 				*color = c.ColorMap[rawPixel]
 			}
-			rawPixels[int(y)*int(rect.Width)+int(x)]=uint32(color.B)<<16 | uint32(color.G)<<8 | uint32(color.R)
+			rawPixels[int(y)*int(rect.Width)+int(x)] = uint32(0xff)<<24 | uint32(color.B)<<16 | uint32(color.G)<<8 | uint32(color.R)
 			//fmt.Printf("%x %x",rawPixel,rawPixels[int(y)*int(rect.Width)+int(x)])
 		}
 	}
 
-	return &RawEncoding{colors,rawPixels}, nil
+	return &RawEncoding{colors, rawPixels}, nil
 }
+
 // ZlibEncoding is raw pixel data sent by the server compressed by Zlib.
 //
 // A single Zlib stream is created. There is only a single header for a framebuffer request response.
 type ZlibEncoding struct {
-	Colors  []Color
-	RawPixel[] uint32
-	ZStream *bytes.Buffer
-	ZReader io.ReadCloser
+	Colors   []Color
+	RawPixel []uint32
+	ZStream  *bytes.Buffer
+	ZReader  io.ReadCloser
 }
 
 func (*ZlibEncoding) Type() int32 {
@@ -87,6 +113,7 @@ func (*ZlibEncoding) Type() int32 {
 }
 
 func (ze *ZlibEncoding) Read(c *ClientConn, rect *Rectangle, r io.Reader) (Encoding, error) {
+	//fmt.Println("ZlibEncoding")
 	bytesPerPixel := c.PixelFormat.BPP / 8
 	pixelBytes := make([]uint8, bytesPerPixel)
 
@@ -98,7 +125,6 @@ func (ze *ZlibEncoding) Read(c *ClientConn, rect *Rectangle, r io.Reader) (Encod
 	// Format
 	// 4 bytes        | uint32 | length
 	// 'length' bytes | []byte | zlibData
-
 	// Read zlib length
 	var zipLength uint32
 	err := binary.Read(r, binary.BigEndian, &zipLength)
@@ -146,7 +172,7 @@ func (ze *ZlibEncoding) Read(c *ClientConn, rect *Rectangle, r io.Reader) (Encod
 	colorReader := bytes.NewReader(colorBytes)
 
 	colors := make([]Color, int(rect.Height)*int(rect.Width))
-	rawPixels:=make([]uint32,int(rect.Height)*int(rect.Width))
+	rawPixels := make([]uint32, int(rect.Height)*int(rect.Width))
 	for y := uint16(0); y < rect.Height; y++ {
 		for x := uint16(0); x < rect.Width; x++ {
 			if _, err := io.ReadFull(colorReader, pixelBytes); err != nil {
@@ -167,14 +193,19 @@ func (ze *ZlibEncoding) Read(c *ClientConn, rect *Rectangle, r io.Reader) (Encod
 				color.R = uint16((rawPixel >> c.PixelFormat.RedShift) & uint32(c.PixelFormat.RedMax))
 				color.G = uint16((rawPixel >> c.PixelFormat.GreenShift) & uint32(c.PixelFormat.GreenMax))
 				color.B = uint16((rawPixel >> c.PixelFormat.BlueShift) & uint32(c.PixelFormat.BlueMax))
+				if c.PixelFormat.BPP == 16 {
+					color.B = color.B<<3 | color.B>>2
+					color.G = color.G<<2 | color.G>>2
+					color.R = color.R<<3 | color.R>>2
+				}
 			} else {
 				*color = c.ColorMap[rawPixel]
 			}
-			rawPixels[int(y)*int(rect.Width)+int(x)]=uint32(color.B)<<16 | uint32(color.G)<<8 | uint32(color.R)
+			rawPixels[int(y)*int(rect.Width)+int(x)] = uint32(0xff)<<24 | uint32(color.B)<<16 | uint32(color.G)<<8 | uint32(color.R)
 		}
 	}
 
-	return &ZlibEncoding{Colors: colors,RawPixel: rawPixels}, nil
+	return &ZlibEncoding{Colors: colors, RawPixel: rawPixels}, nil
 }
 
 func (ze *ZlibEncoding) Close() {
@@ -183,4 +214,4 @@ func (ze *ZlibEncoding) Close() {
 		ze.ZReader.Close()
 		ze.ZReader = nil
 	}
-}
+}

pkg/driver/vncdriver/vncdriver.go

@@ -5,13 +5,14 @@ import (
 	"context"
 	"encoding/binary"
 	"fmt"
-	"github.com/pion/mediadevices/pkg/driver/vncdriver/vnc"
 	"image"
 	"io"
 	"net"
 	"sync"
 	"time"
 
+	"github.com/pion/mediadevices/pkg/driver/vncdriver/vnc"
+
 	"github.com/pion/mediadevices/pkg/frame"
 	"github.com/pion/mediadevices/pkg/io/video"
 	"github.com/pion/mediadevices/pkg/prop"
@@ -32,12 +33,12 @@ func NewVnc(vncAddr string) *vncDevice {
 	return &vncDevice{vncAddr: vncAddr}
 }
 func (d *vncDevice) PointerEvent(mask uint8, x, y uint16) {
-	if d.vClient!=nil{
+	if d.vClient != nil {
 		d.vClient.PointerEvent(vnc.ButtonMask(mask), x, y)
 	}
 }
 func (d *vncDevice) KeyEvent(keysym uint32, down bool) {
-	if d.vClient!=nil {
+	if d.vClient != nil {
 		d.vClient.KeyEvent(keysym, down)
 	}
 }
@@ -49,7 +50,10 @@ func (d *vncDevice) Open() error {
 	d.closed = ctx.Done()
 	d.cancel = cancel
 	msg := make(chan vnc.ServerMessage, 1)
+	//auth:=new(vnc.PasswordAuth)
+	//auth.Password="####"
 	conf := vnc.ClientConfig{
+		//Auth: []vnc.ClientAuth{auth},
 		ServerMessageCh: msg,
 		Exclusive:       false,
 	}
@@ -66,6 +70,7 @@ func (d *vncDevice) Open() error {
 	d.vClient.SetEncodings([]vnc.Encoding{
 		&vnc.ZlibEncoding{},
 		&vnc.RawEncoding{},
+		&vnc.CursorEncoding{},
 	})
 	d.w = int(d.vClient.FrameBufferWidth)
 	d.h = int(d.vClient.FrameBufferHeight)
@@ -89,6 +94,9 @@ func (d *vncDevice) Open() error {
 					for _, rect := range t.Rectangles {
 						var pix []uint32
 						switch t := rect.Enc.(type) {
+						case *vnc.CursorEncoding:
+							//ignore remote cursor messages
+							continue
 						case *vnc.RawEncoding:
 							pix = t.RawPixel
 						case *vnc.ZlibEncoding:
@@ -97,19 +105,16 @@ func (d *vncDevice) Open() error {
 						for y := int(rect.Y); y < int(rect.Height+rect.Y); y++ {
 							for x := int(rect.X); x < int(rect.Width+rect.X); x++ {
 								binary.LittleEndian.PutUint32(d.rawPixel[(y*d.w+x)*4:], pix[(y-int(rect.Y))*int(rect.Width)+(x-int(rect.X))])
-								//BigEndian
 							}
 						}
 
 					}
-					//time.Sleep(33 * time.Millisecond)
 					d.vClient.FramebufferUpdateRequest(true, 0, 0, uint16(d.w), uint16(d.h))
 					break
 				default:
 
 				}
 			case <-time.After(10 * time.Second):
-				//fmt.Println("Timeout FramebufferUpdate")
 				if d.vClient.FramebufferUpdateRequest(true, 0, 0, uint16(d.w), uint16(d.h)) != nil {
 					d.cancel()
 					return
@@ -137,13 +142,12 @@ func (d *vncDevice) Close() error {
 
 func (d *vncDevice) VideoRecord(p prop.Media) (video.Reader, error) {
 	if p.FrameRate == 0 {
-		p.FrameRate = 15
+		p.FrameRate = 30
 	}
 
 	tick := time.NewTicker(time.Duration(float32(time.Second) / p.FrameRate))
 	d.tick = tick
 	closed := d.closed
-	pixs := make([]byte, d.h*d.w*4)
 	r := video.ReaderFunc(func() (image.Image, func(), error) {
 		select {
 		case <-closed:
@@ -153,9 +157,8 @@ func (d *vncDevice) VideoRecord(p prop.Media) (video.Reader, error) {
 		}
 
 		<-tick.C
-		copy(pixs, d.rawPixel)
 		return &image.RGBA{
-			Pix:    pixs,
+			Pix:    d.rawPixel,
 			Stride: 4,
 			Rect:   image.Rect(0, 0, d.w, d.h),
 		}, func() {}, nil

pkg/prop/prop.go

@@ -145,14 +145,17 @@ func (p *MediaConstraints) FitnessDistance(o Media) (float64, bool) {
 	cmps.add(p.Width, o.Width)
 	cmps.add(p.Height, o.Height)
 	cmps.add(p.FrameFormat, o.FrameFormat)
-	cmps.add(p.FrameRate, o.FrameRate)
+	// The next line is comment out for now to not include framerate in the fitness function.
+	// As camera.Properties does not have access to the list of available framerate at the moment,
+	// no driver can be matched with a framerate constraint.
+	// Note this also affect screen caputre as screen.Properties does not fill in the Framerate field.
+	// cmps.add(p.FrameRate, o.FrameRate)
 	cmps.add(p.SampleRate, o.SampleRate)
 	cmps.add(p.Latency, o.Latency)
 	cmps.add(p.ChannelCount, o.ChannelCount)
 	cmps.add(p.IsBigEndian, o.IsBigEndian)
 	cmps.add(p.IsFloat, o.IsFloat)
 	cmps.add(p.IsInterleaved, o.IsInterleaved)
-
 	return cmps.fitnessDistance()
 }