Windows camera driver fixes

pkg/driver/camera/camera_windows.cpp

@@ -2,6 +2,8 @@
 #include <unistd.h>
 
 #include <dshow.h>
+#include <amvideo.h>
+#include <dvdmedia.h>
 #include <qedit.h>
 #include <mmsystem.h>
 
@@ -119,7 +121,7 @@ int freeCameraList(cameraList* list, const char** errstr)
     }
     delete list->name;
   }
-  return 1;
+  return 0;
 }
 
 // selectCamera stores pointer to the selected device IMoniker* according to the configs in camera*.
@@ -175,7 +177,8 @@ int selectCamera(camera* cam, IMoniker** monikerSelected, const char** errstr)
 fail:
   safeRelease(&sysDevEnum);
   safeRelease(&enumMon);
-  return 1;
+  // Signal failure. Callers treat non-zero as success.
+  return 0;
 }
 
 // listResolution stores list of the device to camera*.
@@ -195,7 +198,17 @@ int listResolution(camera* cam, const char** errstr)
     goto fail;
   }
 
-  moniker->BindToObject(0, 0, IID_IBaseFilter, (void**)&captureFilter);
+  if (moniker == nullptr)
+  {
+    *errstr = errEnumDevice;
+    goto fail;
+  }
+
+  if (FAILED(moniker->BindToObject(0, 0, IID_IBaseFilter, (void**)&captureFilter)))
+  {
+    *errstr = errEnumDevice;
+    goto fail;
+  }
   safeRelease(&moniker);
 
   src = getPin(captureFilter, PINDIR_OUTPUT);
@@ -232,14 +245,40 @@ int listResolution(camera* cam, const char** errstr)
         continue;
 
       if (mediaType->majortype != MEDIATYPE_Video ||
-          mediaType->formattype != FORMAT_VideoInfo ||
           mediaType->pbFormat == nullptr)
         continue;
 
-      VIDEOINFOHEADER* videoInfoHdr = (VIDEOINFOHEADER*)mediaType->pbFormat;
-      cam->props[iProp].width = videoInfoHdr->bmiHeader.biWidth;
-      cam->props[iProp].height = videoInfoHdr->bmiHeader.biHeight;
-      cam->props[iProp].fcc = videoInfoHdr->bmiHeader.biCompression;
+      int width = 0;
+      int height = 0;
+      uint32_t fcc = 0;
+      if (mediaType->formattype == FORMAT_VideoInfo)
+      {
+        VIDEOINFOHEADER* videoInfoHdr = (VIDEOINFOHEADER*)mediaType->pbFormat;
+        width = videoInfoHdr->bmiHeader.biWidth;
+        height = videoInfoHdr->bmiHeader.biHeight;
+        fcc = (uint32_t)videoInfoHdr->bmiHeader.biCompression;
+      }
+      else if (mediaType->formattype == FORMAT_VideoInfo2)
+      {
+        VIDEOINFOHEADER2* videoInfoHdr = (VIDEOINFOHEADER2*)mediaType->pbFormat;
+        width = videoInfoHdr->bmiHeader.biWidth;
+        height = videoInfoHdr->bmiHeader.biHeight;
+        fcc = (uint32_t)videoInfoHdr->bmiHeader.biCompression;
+      }
+      else
+      {
+        continue;
+      }
+
+      // biHeight can be negative (top-down). Normalize to a positive size.
+      if (width < 0)
+        width = -width;
+      if (height < 0)
+        height = -height;
+
+      cam->props[iProp].width = width;
+      cam->props[iProp].height = height;
+      cam->props[iProp].fcc = fcc;
       iProp++;
     }
     cam->numProps = iProp;
@@ -283,7 +322,18 @@ int openCamera(camera* cam, const char** errstr)
   {
     goto fail;
   }
-  moniker->BindToObject(0, 0, IID_IBaseFilter, (void**)&captureFilter);
+
+  if (moniker == nullptr)
+  {
+    *errstr = errEnumDevice;
+    goto fail;
+  }
+
+  if (FAILED(moniker->BindToObject(0, 0, IID_IBaseFilter, (void**)&captureFilter)))
+  {
+    *errstr = errEnumDevice;
+    goto fail;
+  }
   safeRelease(&moniker);
 
   if (FAILED(CoCreateInstance(

pkg/driver/camera/camera_windows.go

@@ -10,6 +10,7 @@ import (
 	"image"
 	"io"
 	"sync"
+	"sync/atomic"
 	"unsafe"
 
 	"github.com/pion/mediadevices/pkg/driver"
@@ -25,11 +26,12 @@ var (
 )
 
 type camera struct {
-	name  string
-	cam   *C.camera
-	ch    chan []byte
-	buf   []byte
-	bufGo []byte
+	name   string
+	cam    *C.camera
+	ch     chan []byte
+	buf    []byte
+	bufGo  []byte
+	closed atomic.Bool
 }
 
 func init() {
@@ -75,6 +77,7 @@ func DestroyObserver() error {
 }
 
 func (c *camera) Open() error {
+	c.closed.Store(false)
 	c.ch = make(chan []byte)
 	c.cam = &C.camera{
 		name: C.CString(c.name),
@@ -96,19 +99,35 @@ func imageCallback(cam uintptr) {
 	if !ok {
 		return
 	}
+	if cb.closed.Load() {
+		return
+	}
 
 	copy(cb.bufGo, cb.buf)
-	cb.ch <- cb.bufGo
+	// The native capture graph may deliver frames concurrently with teardown.
+	// Avoid panicking on send-to-closed-channel, and avoid blocking the callback.
+	defer func() {
+		_ = recover()
+	}()
+	select {
+	case cb.ch <- cb.bufGo:
+	default:
+	}
 }
 
 func (c *camera) Close() error {
+	c.closed.Store(true)
+
 	callbacksMu.Lock()
 	key := uintptr(unsafe.Pointer(c.cam))
 	if _, ok := callbacks[key]; ok {
 		delete(callbacks, key)
 	}
 	callbacksMu.Unlock()
-	close(c.ch)
+	if c.ch != nil {
+		close(c.ch)
+		c.ch = nil
+	}
 
 	if c.cam != nil {
 		C.free(unsafe.Pointer(c.cam.name))
@@ -158,20 +177,22 @@ func (c *camera) Properties() []prop.Media {
 	properties := []prop.Media{}
 	for i := 0; i < int(c.cam.numProps); i++ {
 		p := C.getProp(c.cam, C.int(i))
-		// TODO: support other FOURCC
-		if p.fcc == fourccYUY2 {
-			properties = append(properties, prop.Media{
-				Video: prop.Video{
-					Width:       int(p.width),
-					Height:      int(p.height),
-					FrameFormat: frame.FormatYUY2,
-				},
-			})
+		w := int(p.width)
+		h := int(p.height)
+		if w <= 0 || h <= 0 {
+			continue
 		}
+		// We request YUY2 from the DirectShow SampleGrabber.
+		// Even if the camera's native caps are MJPG/etc, DirectShow can often insert
+		// a decoder/colorspace converter so the grabber still receives YUY2.
+		properties = append(properties, prop.Media{
+			Video: prop.Video{
+				Width:       w,
+				Height:      h,
+				FrameFormat: frame.FormatYUY2,
+			},
+		})
 	}
+
 	return properties
 }
-
-const (
-	fourccYUY2 = 0x32595559
-)