coriander-devel Mailing List for Coriander

Hi all,

First of all, thanks for the great job.

I have a problem using a USB3.0 camera with Coriander. My camera is from Point Grey (Grasshopper3, GS3-U3-23S6M-C). I use a (not so recent) HP computer, with i5 CPU, and Orico PCI USB3.0 card (with  Fresco Logic FL1009 USB 3.0 Host Controller), and Ubuntu 14.04, freshly installed.

FlyCapture2 works fine : it detects the camera, and can go up to 125fps (the camera is supposed to go to 160fps, with a newer computer I guess). I can start Coriander, using libdc1394 v2. I just had to add the ID of this camera in /USB/control.c before compiling. 
The ID is the following :   
{ 0x1e10, 0x3103 }, // Point Grey Grasshopper 3
This ID was given by lsusb, when the camera is plugged in a USB2.0 port.

Coriander detects the camera : right vendor and model names are given, and the settings (gain, framerate, etc) have the values that I assigned with FlyCapture2 just before starting Coriander (I exited FlyCapture2 before starting Coriander).

But when I press "Display", I get these error messages in the terminal :

libdc1394 error: usb: Bulk transfer 0 failed with code 6
libdc1394 error: usb: Bulk transfer 1 failed with code 6
libdc1394 error: usb: Bulk transfer 2 failed with code 6
libdc1394 error: usb: Bulk transfer 3 failed with code 6
libdc1394 error: usb: Bulk transfer 4 failed with code 6
libdc1394 error: usb: Bulk transfer 5 failed with code 6
libdc1394 error: usb: Bulk transfer 6 failed with code 6
libdc1394 error: usb: Bulk transfer 7 failed with code 6
libdc1394 error: usb: Bulk transfer 8 failed with code 6
libdc1394 error: usb: Bulk transfer 9 failed with code 6

And then Coriander freezes.

Do you have any idea about this "Bulk transfer" failure ? What is code 6 ? What could I do ?

Thanks for you help.
Maxime Deforet

Hopefully this works!

(As an aside, in the meantime, I've also gotten the mpeg compression
working again.)

Patches for thread_v4l.c and thread_v4l.h

diff -ur coriander/src/thread_v4l.c coriander-v42l/src/thread_v4l.c
--- coriander/src/thread_v4l.c 2014-01-02 12:50:47.485932724 -0600
+++ coriander-v42l/src/thread_v4l.c 2014-01-02 12:51:09.629932142 -0600
@@ -18,6 +18,7 @@
  */

 #include "coriander.h"
+#include <stdio.h>

 gint
 V4lStartThread(camera_t* cam)
@@ -105,8 +106,10 @@
   v4lthread_info_t *info=NULL;
   long int skip_counter;
   float tmp;
-  struct video_picture p;
-
+  //struct video_picture p;
+
+  struct v4l2_format pf;
+
   v4l_service=(chain_t*)arg;
   pthread_mutex_lock(&v4l_service->mutex_data);
   info=(v4lthread_info_t*)v4l_service->data;
@@ -142,15 +145,40 @@

  /* IF we have mono data then set V4L for mono(grey) output */
  /* Only do this ONCE before writing the first frame */
- if (((v4l_service->current_buffer->frame.color_coding ==
DC1394_COLOR_CODING_MONO8) ||
-     (v4l_service->current_buffer->frame.color_coding ==
DC1394_COLOR_CODING_RAW8)) && v4l_service->processed_frames==0) {
-  Warning("Setting V4L device to GREY palette");
-  if (ioctl(info->v4l_dev,VIDIOCGPICT,&p) < 0)
-    Error("ioctl(VIDIOCGPICT) error");
+ if ((v4l_service->current_buffer->frame.color_coding ==
DC1394_COLOR_CODING_MONO8) && v4l_service->processed_frames==0) {
+  Warning("Setting V4L device to GREY pix_fmt");
+          pf.type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
+  if (ioctl(info->v4l_dev,VIDIOC_G_FMT,&pf) < 0)
+    Error("ioctl(VIDIOC_G_FMT) error");
   else {
-    p.palette = VIDEO_PALETTE_GREY;
-    if (ioctl(info->v4l_dev,VIDIOCSPICT,&p) < 0)
-      Error("ioctl(VIDIOCSPICT) Error");
+    pf.fmt.pix.pixelformat = V4L2_PIX_FMT_GREY;
+            pf.fmt.pix.sizeimage =
v4l_service->current_buffer->frame.size[0]*v4l_service->current_buffer->frame.size[1];
+    if (ioctl(info->v4l_dev,VIDIOC_S_FMT,&pf) < 0)
+      Error("ioctl(VIDIOC_S_FMT) Error");
+  }
+ }
+ else if ((v4l_service->current_buffer->frame.color_coding ==
DC1394_COLOR_CODING_RAW8) && v4l_service->processed_frames==0) {
+  Warning("Setting V4L device to RGGB pix_fmt");
+          pf.type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
+  if (ioctl(info->v4l_dev,VIDIOC_G_FMT,&pf) < 0)
+    Error("ioctl(VIDIOC_G_FMT) error");
+  else {
+    pf.fmt.pix.pixelformat = V4L2_PIX_FMT_SRGGB8;
+            pf.fmt.pix.sizeimage =
v4l_service->current_buffer->frame.size[0]*v4l_service->current_buffer->frame.size[1];
+    if (ioctl(info->v4l_dev,VIDIOC_S_FMT,&pf) < 0)
+      Error("ioctl(VIDIOC_S_FMT) Error");
+  }
+ }
+ else if (v4l_service->processed_frames==0) { // Default is RGB24
+  Warning("Setting V4L device to RGB24 pix_fmt");
+          pf.type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
+  if (ioctl(info->v4l_dev,VIDIOC_G_FMT,&pf) < 0)
+    Error("ioctl(VIDIOC_G_FMT) error");
+  else {
+    pf.fmt.pix.pixelformat = V4L2_PIX_FMT_RGB24;
+            pf.fmt.pix.sizeimage =
v4l_service->current_buffer->frame.size[0]*v4l_service->current_buffer->frame.size[1]*3;
+    if (ioctl(info->v4l_dev,VIDIOC_S_FMT,&pf) < 0)
+      Error("ioctl(VIDIOC_S_FMT) Error");
   }
  }

@@ -158,7 +186,7 @@
  if ((v4l_service->current_buffer->frame.color_coding !=
DC1394_COLOR_CODING_MONO8) &&
     (v4l_service->current_buffer->frame.color_coding !=
DC1394_COLOR_CODING_RAW8)) {
   convert_to_rgb(&v4l_service->current_buffer->frame, &info->frame);
-  swap_rb(info->frame.image,
v4l_service->current_buffer->frame.size[0]*v4l_service->current_buffer->frame.size[1]*3);
+  //swap_rb(info->frame.image,
v4l_service->current_buffer->frame.size[0]*v4l_service->current_buffer->frame.size[1]*3);
  }

  if (v4l_service->current_buffer->frame.size[0]!=-1) {
@@ -257,6 +285,7 @@
 {

   v4lthread_info_t *info;
+
   info=(v4lthread_info_t*)v4l_service->data;

   // if some parameters changed, we need to re-allocate the local
buffers and restart the v4l
@@ -266,23 +295,20 @@
     // STOPING THE PIPE MIGHT BE NECESSARY HERE

     // "start pipe"
-    if (ioctl (info->v4l_dev, VIDIOCGCAP, &info->vid_caps) == -1) {
-      perror ("ioctl (VIDIOCGCAP)");
+    if (ioctl (info->v4l_dev, VIDIOC_QUERYCAP, &info->vid_caps) == -1) {
+      perror ("ioctl (VIDIOC_QUERYCAP)");
     }
-    if (ioctl (info->v4l_dev, VIDIOCGPICT, &info->vid_pic)== -1) {
-      perror ("ioctl VIDIOCGPICT");
-    }
-    info->vid_pic.palette = VIDEO_PALETTE_RGB24;
-    if (ioctl (info->v4l_dev, VIDIOCSPICT, &info->vid_pic)== -1) {
-      perror ("ioctl VIDIOCSPICT");
-    }
-    if (ioctl (info->v4l_dev, VIDIOCGWIN, &info->vid_win)== -1) {
-      perror ("ioctl VIDIOCGWIN");
+
+    info->vid_format.type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
+    if (ioctl (info->v4l_dev,VIDIOC_G_FMT,&info->vid_format) == -1) {
+      perror ("ioctl(VIDIOC_G_FMT)");
     }
-    info->vid_win.width=v4l_service->current_buffer->frame.size[0];
-    info->vid_win.height=v4l_service->current_buffer->frame.size[1];
-    if (ioctl (info->v4l_dev, VIDIOCSWIN, &info->vid_win)== -1) {
-      perror ("ioctl VIDIOCSWIN");
+    info->vid_format.type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
+    info->vid_format.fmt.pix.pixelformat = V4L2_PIX_FMT_RGB24; //
default-changes before data gets sent out
+    info->vid_format.fmt.pix.width=v4l_service->current_buffer->frame.size[0];
+    info->vid_format.fmt.pix.height=v4l_service->current_buffer->frame.size[1];
+    if (ioctl (info->v4l_dev, VIDIOC_S_FMT, &info->vid_format)== -1) {
+      perror ("ioctl VIDIOC_S_FMT");
     }
   }

diff -ur coriander/src/thread_v4l.h coriander-v42l/src/thread_v4l.h
--- coriander/src/thread_v4l.h 2014-01-02 12:50:47.485932724 -0600
+++ coriander-v42l/src/thread_v4l.h 2014-01-02 12:51:09.629932142 -0600
@@ -20,6 +20,8 @@
 #ifndef __THREAD_V4L_H__
 #define __THREAD_V4L_H__

+#include <linux/videodev2.h>
+
 typedef struct
 {
   //long int                period;
@@ -30,9 +32,12 @@
   dc1394video_frame_t frame;
   //unsigned char *v4l_buffer;

-  struct video_capability vid_caps;
-  struct video_window vid_win;
-  struct video_picture vid_pic;
+  //struct video_capability vid_caps;
+  //struct video_window vid_win;
+  //struct video_picture vid_pic;
+
+  struct v4l2_capability vid_caps;
+  struct v4l2_format vid_format;

 } v4lthread_info_t;

– caleb


On Wed, Jan 1, 2014 at 9:54 PM, Damien Douxchamps <dd...@do...> wrote:
> Hi Caleb,
>
> On Tue, 2013-12-17 at 11:07 -0600, Caleb Kemere wrote:
>> Hey folks,
>>
>> With my Ubuntu 12.04 system, I had trouble finding a reliable source
>> for a v4lloopback pacakge, so I modified the coriander code to work
>> with v4l2loopback. I'm too lazy to figure out to diff the changes in
>> thread-v4l.*, so I created a github repository for my modified version
>> here: https://github.com/ckemere/coriander-v42l
>
> It's been a long time, but I don't remember seeing a v4l2loopback
> package. This is good news!
>
>> If there's positive feedback, I suppose I can try to figure out the
>> diff/patch syntax...
>
> Try this:
>
> diff -ur original_directory/ my_new_directory/ > v4l2.patch
>
> then share the v4l2.patch file.
>
> Thanks,
>
> --
> Damien 高原 Douxchamps   http://damien.douxchamps.net/
>

Hopefully useful! Note that the raw / mono formats will probably not
work. If you convert to color before enabling the V4L service, things
seem to be happy.

(As an aside, in the meantime, I've also gotten the mpeg compression
working again. The most recent version on my github.com/ckemere
repository also includes that.)

Patches for thread_v4l.c and thread_v4l.h:

diff -ur coriander/src/thread_v4l.c coriander-v42l/src/thread_v4l.c
--- coriander/src/thread_v4l.c 2014-01-02 12:50:47.485932724 -0600
+++ coriander-v42l/src/thread_v4l.c 2014-01-02 12:51:09.629932142 -0600
@@ -18,6 +18,7 @@
  */

 #include "coriander.h"
+#include <stdio.h>

 gint
 V4lStartThread(camera_t* cam)
@@ -105,8 +106,10 @@
   v4lthread_info_t *info=NULL;
   long int skip_counter;
   float tmp;
-  struct video_picture p;
-
+  //struct video_picture p;
+
+  struct v4l2_format pf;
+
   v4l_service=(chain_t*)arg;
   pthread_mutex_lock(&v4l_service->mutex_data);
   info=(v4lthread_info_t*)v4l_service->data;
@@ -142,15 +145,40 @@

  /* IF we have mono data then set V4L for mono(grey) output */
  /* Only do this ONCE before writing the first frame */
- if (((v4l_service->current_buffer->frame.color_coding ==
DC1394_COLOR_CODING_MONO8) ||
-     (v4l_service->current_buffer->frame.color_coding ==
DC1394_COLOR_CODING_RAW8)) && v4l_service->processed_frames==0) {
-  Warning("Setting V4L device to GREY palette");
-  if (ioctl(info->v4l_dev,VIDIOCGPICT,&p) < 0)
-    Error("ioctl(VIDIOCGPICT) error");
+ if ((v4l_service->current_buffer->frame.color_coding ==
DC1394_COLOR_CODING_MONO8) && v4l_service->processed_frames==0) {
+  Warning("Setting V4L device to GREY pix_fmt");
+          pf.type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
+  if (ioctl(info->v4l_dev,VIDIOC_G_FMT,&pf) < 0)
+    Error("ioctl(VIDIOC_G_FMT) error");
   else {
-    p.palette = VIDEO_PALETTE_GREY;
-    if (ioctl(info->v4l_dev,VIDIOCSPICT,&p) < 0)
-      Error("ioctl(VIDIOCSPICT) Error");
+    pf.fmt.pix.pixelformat = V4L2_PIX_FMT_GREY;
+            pf.fmt.pix.sizeimage =
v4l_service->current_buffer->frame.size[0]*v4l_service->current_buffer->frame.size[1];
+    if (ioctl(info->v4l_dev,VIDIOC_S_FMT,&pf) < 0)
+      Error("ioctl(VIDIOC_S_FMT) Error");
+  }
+ }
+ else if ((v4l_service->current_buffer->frame.color_coding ==
DC1394_COLOR_CODING_RAW8) && v4l_service->processed_frames==0) {
+  Warning("Setting V4L device to RGGB pix_fmt");
+          pf.type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
+  if (ioctl(info->v4l_dev,VIDIOC_G_FMT,&pf) < 0)
+    Error("ioctl(VIDIOC_G_FMT) error");
+  else {
+    pf.fmt.pix.pixelformat = V4L2_PIX_FMT_SRGGB8;
+            pf.fmt.pix.sizeimage =
v4l_service->current_buffer->frame.size[0]*v4l_service->current_buffer->frame.size[1];
+    if (ioctl(info->v4l_dev,VIDIOC_S_FMT,&pf) < 0)
+      Error("ioctl(VIDIOC_S_FMT) Error");
+  }
+ }
+ else if (v4l_service->processed_frames==0) { // Default is RGB24
+  Warning("Setting V4L device to RGB24 pix_fmt");
+          pf.type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
+  if (ioctl(info->v4l_dev,VIDIOC_G_FMT,&pf) < 0)
+    Error("ioctl(VIDIOC_G_FMT) error");
+  else {
+    pf.fmt.pix.pixelformat = V4L2_PIX_FMT_RGB24;
+            pf.fmt.pix.sizeimage =
v4l_service->current_buffer->frame.size[0]*v4l_service->current_buffer->frame.size[1]*3;
+    if (ioctl(info->v4l_dev,VIDIOC_S_FMT,&pf) < 0)
+      Error("ioctl(VIDIOC_S_FMT) Error");
   }
  }

@@ -158,7 +186,7 @@
  if ((v4l_service->current_buffer->frame.color_coding !=
DC1394_COLOR_CODING_MONO8) &&
     (v4l_service->current_buffer->frame.color_coding !=
DC1394_COLOR_CODING_RAW8)) {
   convert_to_rgb(&v4l_service->current_buffer->frame, &info->frame);
-  swap_rb(info->frame.image,
v4l_service->current_buffer->frame.size[0]*v4l_service->current_buffer->frame.size[1]*3);
+  //swap_rb(info->frame.image,
v4l_service->current_buffer->frame.size[0]*v4l_service->current_buffer->frame.size[1]*3);
  }

  if (v4l_service->current_buffer->frame.size[0]!=-1) {
@@ -257,6 +285,7 @@
 {

   v4lthread_info_t *info;
+
   info=(v4lthread_info_t*)v4l_service->data;

   // if some parameters changed, we need to re-allocate the local
buffers and restart the v4l
@@ -266,23 +295,20 @@
     // STOPING THE PIPE MIGHT BE NECESSARY HERE

     // "start pipe"
-    if (ioctl (info->v4l_dev, VIDIOCGCAP, &info->vid_caps) == -1) {
-      perror ("ioctl (VIDIOCGCAP)");
+    if (ioctl (info->v4l_dev, VIDIOC_QUERYCAP, &info->vid_caps) == -1) {
+      perror ("ioctl (VIDIOC_QUERYCAP)");
     }
-    if (ioctl (info->v4l_dev, VIDIOCGPICT, &info->vid_pic)== -1) {
-      perror ("ioctl VIDIOCGPICT");
-    }
-    info->vid_pic.palette = VIDEO_PALETTE_RGB24;
-    if (ioctl (info->v4l_dev, VIDIOCSPICT, &info->vid_pic)== -1) {
-      perror ("ioctl VIDIOCSPICT");
-    }
-    if (ioctl (info->v4l_dev, VIDIOCGWIN, &info->vid_win)== -1) {
-      perror ("ioctl VIDIOCGWIN");
+
+    info->vid_format.type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
+    if (ioctl (info->v4l_dev,VIDIOC_G_FMT,&info->vid_format) == -1) {
+      perror ("ioctl(VIDIOC_G_FMT)");
     }
-    info->vid_win.width=v4l_service->current_buffer->frame.size[0];
-    info->vid_win.height=v4l_service->current_buffer->frame.size[1];
-    if (ioctl (info->v4l_dev, VIDIOCSWIN, &info->vid_win)== -1) {
-      perror ("ioctl VIDIOCSWIN");
+    info->vid_format.type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
+    info->vid_format.fmt.pix.pixelformat = V4L2_PIX_FMT_RGB24; //
default-changes before data gets sent out
+    info->vid_format.fmt.pix.width=v4l_service->current_buffer->frame.size[0];
+    info->vid_format.fmt.pix.height=v4l_service->current_buffer->frame.size[1];
+    if (ioctl (info->v4l_dev, VIDIOC_S_FMT, &info->vid_format)== -1) {
+      perror ("ioctl VIDIOC_S_FMT");
     }
   }

diff -ur coriander/src/thread_v4l.h coriander-v42l/src/thread_v4l.h
--- coriander/src/thread_v4l.h 2014-01-02 12:50:47.485932724 -0600
+++ coriander-v42l/src/thread_v4l.h 2014-01-02 12:51:09.629932142 -0600
@@ -20,6 +20,8 @@
 #ifndef __THREAD_V4L_H__
 #define __THREAD_V4L_H__

+#include <linux/videodev2.h>
+
 typedef struct
 {
   //long int                period;
@@ -30,9 +32,12 @@
   dc1394video_frame_t frame;
   //unsigned char *v4l_buffer;

-  struct video_capability vid_caps;
-  struct video_window vid_win;
-  struct video_picture vid_pic;
+  //struct video_capability vid_caps;
+  //struct video_window vid_win;
+  //struct video_picture vid_pic;
+
+  struct v4l2_capability vid_caps;
+  struct v4l2_format vid_format;

 } v4lthread_info_t;

– caleb


On Wed, Jan 1, 2014 at 9:54 PM, Damien Douxchamps <dd...@do...> wrote:
> Hi Caleb,
>
> On Tue, 2013-12-17 at 11:07 -0600, Caleb Kemere wrote:
>> Hey folks,
>>
>> With my Ubuntu 12.04 system, I had trouble finding a reliable source
>> for a v4lloopback pacakge, so I modified the coriander code to work
>> with v4l2loopback. I'm too lazy to figure out to diff the changes in
>> thread-v4l.*, so I created a github repository for my modified version
>> here: https://github.com/ckemere/coriander-v42l
>
> It's been a long time, but I don't remember seeing a v4l2loopback
> package. This is good news!
>
>> If there's positive feedback, I suppose I can try to figure out the
>> diff/patch syntax...
>
> Try this:
>
> diff -ur original_directory/ my_new_directory/ > v4l2.patch
>
> then share the v4l2.patch file.
>
> Thanks,
>
> --
> Damien 高原 Douxchamps   http://damien.douxchamps.net/
>

Hi Caleb,

On Tue, 2013-12-17 at 11:07 -0600, Caleb Kemere wrote: 
> Hey folks,
> 
> With my Ubuntu 12.04 system, I had trouble finding a reliable source
> for a v4lloopback pacakge, so I modified the coriander code to work
> with v4l2loopback. I'm too lazy to figure out to diff the changes in
> thread-v4l.*, so I created a github repository for my modified version
> here: https://github.com/ckemere/coriander-v42l

It's been a long time, but I don't remember seeing a v4l2loopback
package. This is good news!

> If there's positive feedback, I suppose I can try to figure out the
> diff/patch syntax...

Try this:

diff -ur original_directory/ my_new_directory/ > v4l2.patch

then share the v4l2.patch file.

Thanks,

-- 
Damien 高原 Douxchamps   http://damien.douxchamps.net/

Hey folks,

With my Ubuntu 12.04 system, I had trouble finding a reliable source
for a v4lloopback pacakge, so I modified the coriander code to work
with v4l2loopback. I'm too lazy to figure out to diff the changes in
thread-v4l.*, so I created a github repository for my modified version
here: https://github.com/ckemere/coriander-v42l

If there's positive feedback, I suppose I can try to figure out the
diff/patch syntax...

– caleb

Hi all,

It's been a long time since the last release of Coriander! There's
nothing ground breaking, but some overdue minor updates were necessary. 

The only new feature is TIFF image output (16bit too!).

On feature that was removed is the RAM buffering. These days It's
trivial to use /dev/shm for this purpose so the code was binned.

Cheers,

Damien


-- 
Damien 高原 Douxchamps   http://damien.douxchamps.net/

From:	Stefan Richter <st...@s5...>
Date:	Mon, 18 Jun 2012 22:18:17 +0200
> Could be that vloopback no longer provides a proc interface, 
> I don't know.  

If the question can not be answered here, is there a 
better place to ask?  For me to unravel this by studying 
the source code will be infeasible.

Thanks,          ... Peter E.

-- 
Telephone 1 360 639 0202.  Bcc: peter at easthope.ca
"http://carnot.yi.org/ "
"http://members.shaw.ca/peasthope/index.html#Itinerary "

On Jun 18 pea...@sh... wrote:
> From:	Stefan Richter <st...@s5...>
> Date:	Mon, 11 Jun 2012 22:50:51 +0200
> > ... look up in /proc/video/vloopback/ which two video devices belong to one
> > pipe.
> 
> peter@joule:~$ ls /proc/v*
> /proc/version  /proc/vmallocinfo  /proc/vmstat
> 
> No /proc/video/?  How should the directory be created?
> Any ideas why creation fails?

I took that from an old webpage.  Could be that vloopback no longer
provides a proc interface, I don't know.  Or it is simply broken already on
the kernel version that you are using.
-- 
Stefan Richter
-=====-===-- -==- =--=-
http://arcgraph.de/sr/

From:	Stefan Richter <st...@s5...>
Date:	Mon, 11 Jun 2012 22:50:51 +0200
> ... look up in /proc/video/vloopback/ which two video devices belong to one
> pipe.

peter@joule:~$ ls /proc/v*
/proc/version  /proc/vmallocinfo  /proc/vmstat

No /proc/video/?  How should the directory be created?
Any ideas why creation fails?

Thanks,          ... Peter E.

-- 
Telephone 1 360 639 0202.  Bcc: peter at easthope.ca
"http://carnot.yi.org/ "
"http://members.shaw.ca/peasthope/index.html#Itinerary "

On Jun 11 pea...@sh... wrote:
> By the way, the Coriander manual has no mention of JuJu or udev and 
> I wonder which of the instructions apply to my Debian Squeeze system.
> 
> peter@joule:/dev$ uname -rv
> 2.6.32-5-686 #1 SMP Sat May 5 01:33:08 UTC 2012
> 
> http://damien.douxchamps.net/ieee1394/coriander/manual/ states,
> "... you should create the proper 1394 devices. There 
> are two devices to create: raw1394 and video1394."

This is outdated.  These devices have been managed by udev for a long
time now.  Nowadays they are replaced by /dev/fw* files though.  However,
since Coriander captures video, everything is alright on that side.

> I would have expected udev to create these devices.  In any case, 
> this is what exists when Coriander is running.
> 
> peter@joule:/dev$ ls -ld *1394*
> ls: cannot access *1394*: No such file or directory
> 
> peter@joule:/dev$ ls -ld *vid*
> crw-rw----+ 1 root video 81, 0 Jun 10 16:20 video0
> crw-rw----+ 1 root video 81, 1 Jun 10 16:20 video1
> crw-rw----+ 1 root video 81, 2 Jun 10 16:20 video2
> crw-rw----+ 1 root video 81, 3 Jun 10 16:20 video3
> 
> Thanks for any ideas,          ... Peter E.

Two or more of these files are certainly backed by vloopback.  I suppose
two by default but four with "modprobe vloopback pipes=2".  Supposedly you
can look up in /proc/video/vloopback/ which two video devices belong to one
pipe.

If "id" says that your account is in the video group, then access
permissions are not the issue.
-- 
Stefan Richter
-=====-===-- -==- -=-==
http://arcgraph.de/sr/

By the way, the Coriander manual has no mention of JuJu or udev and 
I wonder which of the instructions apply to my Debian Squeeze system.

peter@joule:/dev$ uname -rv
2.6.32-5-686 #1 SMP Sat May 5 01:33:08 UTC 2012

http://damien.douxchamps.net/ieee1394/coriander/manual/ states,
"... you should create the proper 1394 devices. There 
are two devices to create: raw1394 and video1394."

I would have expected udev to create these devices.  In any case, 
this is what exists when Coriander is running.

peter@joule:/dev$ ls -ld *1394*
ls: cannot access *1394*: No such file or directory

peter@joule:/dev$ ls -ld *vid*
crw-rw----+ 1 root video 81, 0 Jun 10 16:20 video0
crw-rw----+ 1 root video 81, 1 Jun 10 16:20 video1
crw-rw----+ 1 root video 81, 2 Jun 10 16:20 video2
crw-rw----+ 1 root video 81, 3 Jun 10 16:20 video3

Thanks for any ideas,          ... Peter E.

-- 
Telephone 1 360 639 0202.  Bcc: peter at easthope.ca
"http://carnot.yi.org/ "
"http://members.shaw.ca/peasthope/index.html#Itinerary "

Shelley's photos 
https://picasaweb.google.com/112944318138461249464/May292012?authuser=0&authkey=Gv1sRgCN_c7efw5-rWEg&feat=directlink

Peter's photos
https://picasaweb.google.com/112944318138461249464/May29201202?authuser=0&feat=directlink

From:	Gordon Sinclair <go...@gy...>
Date:	Tue, 05 Jun 2012 21:39:38 +0100
> The only thing i can see is that I always start coriander as root to get
> over any privilege issues. If coriander can not see any cameras try
> starting sudo coriander

For user and for root Coriander  runs and Services > Display gives an image.
No problem there.  The problem is to get the V4L signal.

For interest, I changed the 1394 pci adapter card to a newer Belkin.  hwinfo 
reports the older adapter as well as this one; but "hwinfo --camera" still 
gives nothing.  Can anyone get camera information from hwinfo?  Perhaps 
hwinfo is simply unable to detect a 1394 camera.  

From:	Stefan Richter <st...@s5...>
Date:	Wed, 6 Jun 2012 13:43:17 +0200
> Likewise, ownerships or ACLs on the v4l character device files need to be
> set properly, ...

peter@joule:/dev$ find . -type f -name '*v4l*' -print
peter@joule:/dev$

No /dev file with "v4l" in the name.  What should exist?  
Is it persistent?  Or does it come into existence only when Coriander 
is running the V4L service?

> ... and your user account should be member of the necessary
> group(s) of course.

Which groups?  I don't recognize any group as specific to V4L.  
Should there be a *v4l* group?

Thanks,                   ... Peter E.

-- 
Telephone 1 360 639 0202.  Bcc: peter at easthope.ca
"http://carnot.yi.org/ "
"http://members.shaw.ca/peasthope/index.html#Itinerary "

Shelley's photos 
https://picasaweb.google.com/112944318138461249464/May292012?authuser=0&authkey=Gv1sRgCN_c7efw5-rWEg&feat=directlink

Peter's photos
https://picasaweb.google.com/112944318138461249464/May29201202?authuser=0&feat=directlink

On Jun 05 Gordon Sinclair wrote:
> The only thing i can see is that I always start coriander as root to get
> over any privilege issues. If coriander can not see any cameras try
> starting sudo coriander

Both safer and more practical would be to check why udev doesn't set
useful group ownership or ACLs on the /dev/fw* character device file which
corresponds to the camera, and fix that.

Stock udev has rules for IIDC compliant devices.
https://ieee1394.wiki.kernel.org/index.php/Juju_Migration#Permissions_and_ownership_for_.2Fdev.2Ffw.2A
http://git.kernel.org/?p=linux/hotplug/udev.git;a=blob;f=rules/50-udev-default.rules;h=5ad787fc764594e2999809fee105dde416a3d6e5;hb=HEAD#l48

Likewise, ownerships or ACLs on the v4l character device files need to be
set properly, and your user account should be member of the necessary
group(s) of course.

If it turns out that stock udev rules lack some device IDs, we should
submit a fix to the udev maintainers.
-- 
Stefan Richter
-=====-===-- -==- --==-
http://arcgraph.de/sr/

Gordon,

> If coriander can not see any cameras try
> starting sudo coriander

Will try it this coming weekend.  The camera is at home and I 
am at work in the city.

> ... i only offer you what little help i can be

Much appreciated.  There was no response from a query to 
coriander-user over a yar back.

Regards,           ... P.


-- 
Telephone 1 360 639 0202.  Bcc: peter at easthope.ca
"http://carnot.yi.org/ "
"http://members.shaw.ca/peasthope/index.html#Itinerary "

Dear Peter E.

The only thing i can see is that I always start coriander as root to get
over any privilege issues. If coriander can not see any cameras try
starting sudo coriander

again i can only say i have only just started using the program myself
so i only offer you what little help i can be

thanks

gordon







On Mon, 2012-06-04 at 10:00 -0700, pea...@sh... wrote:
> Gordon and anyone interested,
> 
> Thanks for your instructions.
> 
> > svn co http://www.lavrsen.dk/svn/vloopback/trunk/ vloopback
> > 
> > then compile and install
> 
> I installed from the debian vloopback source package.  Compiled it.
> 
> http://packages.debian.org/source/stable/graphics/vloopback-source
> 
> Date:	Mon, 04 Jun 2012 15:53:37 +0100
> > modprobe vloopback pipes=2
> 
> OK thanks.  I don't recall seeing "pipes=2" mentioned until now.
> 
> peter@joule:~$ lsmod | grep vloop
> vloopback               9842  2
> videodev               25637  3 vloopback
> 
> > run coriander set up the camera ...
> 
> I had to hotplug the camera before Coriander could operate.
> 
> root@joule:/home/peter# hwinfo --all > ./Hardware
> The output file is visible here.
> 
> http://members.shaw.ca/peasthope/Hardware
> 
> The 1394 adapter with TI chip is listed.  No camera.
> 
> > ... and start the v4l service using the default device (/dev/video0)
> 
> Two windows appear.
> 
> "Error: could not start the V4L service."
> 
> "Failed to open V4L device /dev/video1"
> 
> Problem in the Debian vloopback-source package?
> Problem with the old 1394 adapter card with TI chip?
> 
> I can try a different adapter and try compiling from 
> the svn you mentioned.
> 
> Thanks again,                 ... Peter E.
> 
> 

Gordon and anyone interested,

Thanks for your instructions.

> svn co http://www.lavrsen.dk/svn/vloopback/trunk/ vloopback
> 
> then compile and install

I installed from the debian vloopback source package.  Compiled it.

http://packages.debian.org/source/stable/graphics/vloopback-source

Date:	Mon, 04 Jun 2012 15:53:37 +0100
> modprobe vloopback pipes=2

OK thanks.  I don't recall seeing "pipes=2" mentioned until now.

peter@joule:~$ lsmod | grep vloop
vloopback               9842  2
videodev               25637  3 vloopback

> run coriander set up the camera ...

I had to hotplug the camera before Coriander could operate.

root@joule:/home/peter# hwinfo --all > ./Hardware
The output file is visible here.

http://members.shaw.ca/peasthope/Hardware

The 1394 adapter with TI chip is listed.  No camera.

> ... and start the v4l service using the default device (/dev/video0)

Two windows appear.

"Error: could not start the V4L service."

"Failed to open V4L device /dev/video1"

Problem in the Debian vloopback-source package?
Problem with the old 1394 adapter card with TI chip?

I can try a different adapter and try compiling from 
the svn you mentioned.

Thanks again,                 ... Peter E.


-- 
Telephone 1 360 639 0202.  Bcc: peter at easthope.ca
"http://carnot.yi.org/ "
"http://members.shaw.ca/peasthope/index.html#Itinerary "

Dear Peter E.

I do not claim any expertise i have only just started using coriander
and vloopback. But what i have i will share. 

I suggest you get vloopback working its fine on Linux 2.6.38
pre-releases. Download the using subversion

svn co http://www.lavrsen.dk/svn/vloopback/trunk/ vloopback

then compile and install

if you are using One camera you will need two pipes so

modprobe vloopback pipes=2

run coriander set up the camera and start the v4l service using the
default device (/dev/video0)

then run what ever program using the next device ie /dev/video1

should work fine

all the best

gordon



On Sun, 2012-06-03 at 12:30 -0700, pe...@ea... wrote:
> Hi,
> 
> Please excuse this interjection.  Your discussion has piqued 
> my interest.
> 
> About 3 years ago I purchased a Uibrain Fire-i camera and 
> connected it to my  Debian system; at that time Debian Sarge; 
> currently Debian Squeeze with linux-image-2.6.32-5-686.
> vloopback is installed.
> 
> I've never observed any problem with Coriander.  Disappointingly, 
> the camera has never worked for Skype or any other VoIP.
> 
> On May 29 Gordon Sinclair wrote:
> > I have invested in a couple of very nice firewire cams which give me 30
> > fps. I need to get them to work through coriander to my program using
> > the v4l interface. That's when it all went wrong because the V4L1 API
> > was removed in one of the Linux 2.6.38 pre-releases. The
> > <linux/videodev.h> header doesn't exist any more and has been superseded
> > by <linux/videodev2.h>. Thus I cannot build vloopback on my system with
> > the kernel version 2.6.38-15-generic-pae
> 
> This suggests that the problem here may not be my incompetence.  Does every 
> VoIP in Linux require vloopback?  Should vloopback work in this system?
> 
> Any tips welcome, thanks,           ... Peter E.
> 
> 

Hi,

Please excuse this interjection.  Your discussion has piqued 
my interest.

About 3 years ago I purchased a Uibrain Fire-i camera and 
connected it to my  Debian system; at that time Debian Sarge; 
currently Debian Squeeze with linux-image-2.6.32-5-686.
vloopback is installed.

I've never observed any problem with Coriander.  Disappointingly, 
the camera has never worked for Skype or any other VoIP.

On May 29 Gordon Sinclair wrote:
> I have invested in a couple of very nice firewire cams which give me 30
> fps. I need to get them to work through coriander to my program using
> the v4l interface. That's when it all went wrong because the V4L1 API
> was removed in one of the Linux 2.6.38 pre-releases. The
> <linux/videodev.h> header doesn't exist any more and has been superseded
> by <linux/videodev2.h>. Thus I cannot build vloopback on my system with
> the kernel version 2.6.38-15-generic-pae

This suggests that the problem here may not be my incompetence.  Does every 
VoIP in Linux require vloopback?  Should vloopback work in this system?

Any tips welcome, thanks,           ... Peter E.


-- 
Telephone 1 360 639 0202.  Bcc: peter at easthope.ca
"http://carnot.yi.org/ "
"http://members.shaw.ca/peasthope/index.html#Itinerary "

On Jun 03 Gordon Sinclair wrote:
[IIDC->V4L2 driver]
> I agree I think it is the way forward, but as i have 13 mouths of my
> doctorate left and simply do not have the time for now but after i
> graduate i will look at my schedule.

With your time constraints, you should certainly simply install the last
kernel which was still compatible with the old vloopback module.  I suspect
not even a v4l2loopback mode for Coriander will appear soon enough for you,
even if somebody decides to implement it.
-- 
Stefan Richter
-=====-===-- -==- ---==
http://arcgraph.de/sr/

Dear Stefan

I agree I think it is the way forward, but as i have 13 mouths of my
doctorate left and simply do not have the time for now but after i
graduate i will look at my schedule.

all the best

gordon







On Sun, 2012-06-03 at 13:08 +0200, Stefan Richter wrote:
> On May 29 Gordon Sinclair wrote:
> > I have invested in a couple of very nice firewire cams which give me 30
> > fps. I need to get them to work through coriander to my program using
> > the v4l interface. That's when it all went wrong because the V4L1 API
> > was removed in one of the Linux 2.6.38 pre-releases. The
> > <linux/videodev.h> header doesn't exist any more and has been superseded
> > by <linux/videodev2.h>. Thus I cannot build vloopback on my system with
> > the kernel version 2.6.38-15-generic-pae
> > 
> > Is there a branch of coriander which does v4l2loopback?
> > 
> > I have taken a look at the code for vloopback but I don't understand it
> > at all. The only other option is to roll back to a previous kernel.
> 
> Doing FireWire video capture via special low-level IEEE 1394 I/O kernel
> interfaces instead of the V4L/ V4L2 interface has been, and still is, one
> of the big mistakes of IEEE 1394 support on Linux.
> 
> The stop-gap solution for 1394 video via V4L, vloopback, was always
> problematic by
>   (a) being a project outside of the mainline kernel, thus risking
>       breakage at any kernel update,
>   (b) having to rely on a userspace process for DMA buffer queueing.
> The very same problems apply to v4l2loopback of course.
> 
> To me, an IIDC to V4L2 kernel driver (as well as an IEC61883-2/-4 to V4L2
> kernel driver) looks like a much more useful project.  Alas I have far to
> little time for Linux projects myself for the time being to implement such
> a driver.

On May 29 Gordon Sinclair wrote:
> I have invested in a couple of very nice firewire cams which give me 30
> fps. I need to get them to work through coriander to my program using
> the v4l interface. That's when it all went wrong because the V4L1 API
> was removed in one of the Linux 2.6.38 pre-releases. The
> <linux/videodev.h> header doesn't exist any more and has been superseded
> by <linux/videodev2.h>. Thus I cannot build vloopback on my system with
> the kernel version 2.6.38-15-generic-pae
> 
> Is there a branch of coriander which does v4l2loopback?
> 
> I have taken a look at the code for vloopback but I don't understand it
> at all. The only other option is to roll back to a previous kernel.

Doing FireWire video capture via special low-level IEEE 1394 I/O kernel
interfaces instead of the V4L/ V4L2 interface has been, and still is, one
of the big mistakes of IEEE 1394 support on Linux.

The stop-gap solution for 1394 video via V4L, vloopback, was always
problematic by
  (a) being a project outside of the mainline kernel, thus risking
      breakage at any kernel update,
  (b) having to rely on a userspace process for DMA buffer queueing.
The very same problems apply to v4l2loopback of course.

To me, an IIDC to V4L2 kernel driver (as well as an IEC61883-2/-4 to V4L2
kernel driver) looks like a much more useful project.  Alas I have far to
little time for Linux projects myself for the time being to implement such
a driver.
-- 
Stefan Richter
-=====-===-- -==- ---==
http://arcgraph.de/sr/

Dear Developers

I am a doctorate student at the University of Portsmouth.

I have built an eye tracking microscope using two usb web cams to
capture eye moments. Unfortunately I am only getting 8 frames per second
so I think I am missing the very behaviour I need to see.

I have invested in a couple of very nice firewire cams which give me 30
fps. I need to get them to work through coriander to my program using
the v4l interface. That's when it all went wrong because the V4L1 API
was removed in one of the Linux 2.6.38 pre-releases. The
<linux/videodev.h> header doesn't exist any more and has been superseded
by <linux/videodev2.h>. Thus I cannot build vloopback on my system with
the kernel version 2.6.38-15-generic-pae

Is there a branch of coriander which does v4l2loopback?

I have taken a look at the code for vloopback but I don't understand it
at all. The only other option is to roll back to a previous kernel.

Thanks

gordon

On Nov 17 pea...@sh... wrote:
> There should be a definitive way to know whether Coriander is using the 
> "old" 1394 drivers or the Juju drivers.

Coriander just uses libdc1394 for all 1394 I/O, and libdc1394 uses what
the running kernel offers.

> Something in /proc shows this?

/proc/interrupts, /sys/bus/{ieee1394,firewire}/, lspci -v for example...
-- 
Stefan Richter
-=====-==-== =-== =---=
http://arcgraph.de/sr/

Hello Coriander developers,

I am using Coriander in Debian Squeeze with a Unibrain Fire-i camera.
Here are a few suggestions for the documentation.

http://damien.douxchamps.net/ieee1394/cameras/ reports the optics of the 
Fire-i as fixed.  In fact the lens is thread mounted although not removable.  
Focus can be adjusted manually.  The Optics Key doesn't have a notation for 
this configuration.

There should be a definitive way to know whether Coriander is using the 
"old" 1394 drivers or the Juju drivers.  Something in /proc shows this?  
Documention is needed.

http://damien.douxchamps.net/ieee1394/coriander/manual/ and 
"man coriander" should tell which device input is taken from.  Also, when 
vloopback is operating, the documentation should tell which device
a secondary application can obtain output from.

Stefan Richter carefully explained most of this in 
<20111116193303.7d625e33@stein> to coriander-user at 
Date: Wed, 16 Nov 2011 19:33:03 +010.  Incorporation into the Web 
based and traditional manuals would be really helpful for clueless 
novices.

Thanks,                         ... Peter E.



-- 
New telephone 1 360 639 0202.  Old telephone 1 360 450 2132.  
bcc: peasthope at shaw.ca
Work "http://carnot.yi.org/".
"http://members.shaw.ca/peasthope/index.html#Itinerary".

Hi.

Since videodev.h was removed from the linux kernel 2.6.38 and I could
not compile the vloopback that coriander uses/recomends (
http://www.lavrsen.dk/foswiki/bin/view/Motion/VideoFourLinuxLoopbackDevice
) I made it work with the V4L2 (
http://code.google.com/p/v4l2loopback/ ) It seems to work but this was
a quick fix I did.

So anyhow patch attached (this was against git
aec43533bdaf3477ad7660a5683638872801e6c0)

 configure.in     |    2 +-
 src/coriander.h  |    2 +-
 src/thread_v4l.c |   51 ++++++++++++++++++++++++++++-----------------------
 src/thread_v4l.h |    5 ++---
 4 files changed, 32 insertions(+), 28 deletions(-)

diff --git a/configure.in b/configure.in
index 9d758a4..4dfff60 100644
--- a/configure.in
+++ b/configure.in
@@ -4,7 +4,7 @@ dnl AC_INIT(configure.in)
 AC_INIT(coriander, 2.0.1)
 AM_INIT_AUTOMAKE
 AM_MAINTAINER_MODE
-AM_CONFIG_HEADER(config.h)
+dnl AM_CONFIG_HEADER(config.h)

 AC_ISC_POSIX
 AC_PROG_CC
diff --git a/src/coriander.h b/src/coriander.h
index 562005b..0e14ee9 100644
--- a/src/coriander.h
+++ b/src/coriander.h
@@ -59,7 +59,7 @@
 #include <netinet/in.h>

 #include "definitions.h"
-#include "videodev.h"
+#include "linux/videodev2.h"

 // SDL lib
 #ifdef HAVE_SDLLIB
diff --git a/src/thread_v4l.c b/src/thread_v4l.c
index 839cc90..ad14536 100644
--- a/src/thread_v4l.c
+++ b/src/thread_v4l.c
@@ -52,6 +52,8 @@ V4lStartThread(camera_t* cam)

     info->frame.image=NULL;
     info->frame.allocated_image_bytes=0;
+    memset(&info->vid_fmt, 0, sizeof(info->vid_fmt));
+    info->vid_fmt.type = V4L2_BUF_TYPE_VIDEO_OUTPUT;

     // open V4L device
     info->v4l_dev=-1;
@@ -105,8 +107,11 @@ V4lThread(void* arg)
   v4lthread_info_t *info=NULL;
   long int skip_counter;
   float tmp;
-  struct video_picture p;
+  struct v4l2_format fmt;

+  memset(&fmt, 0, sizeof(fmt));
+  fmt.type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
+
   v4l_service=(chain_t*)arg;
   pthread_mutex_lock(&v4l_service->mutex_data);
   info=(v4lthread_info_t*)v4l_service->data;
@@ -127,6 +132,7 @@ V4lThread(void* arg)
   v4l_service->drop_warning = 0;
   v4l_service->processed_frames=0;

+
   while (1) {
     /* Clean cancel handlers */
     pthread_mutex_lock(&info->mutex_cancel);
@@ -145,12 +151,15 @@ V4lThread(void* arg)
 	if (((v4l_service->current_buffer->frame.color_coding ==
DC1394_COLOR_CODING_MONO8) ||
 	     (v4l_service->current_buffer->frame.color_coding ==
DC1394_COLOR_CODING_RAW8)) && v4l_service->processed_frames==0) {
 	  Warning("Setting V4L device to GREY palette");
-	  if (ioctl(info->v4l_dev,VIDIOCGPICT,&p) < 0)
-	    Error("ioctl(VIDIOCGPICT) error");
-	  else {
-	    p.palette = VIDEO_PALETTE_GREY;
-	    if (ioctl(info->v4l_dev,VIDIOCSPICT,&p) < 0)
-	      Error("ioctl(VIDIOCSPICT) Error");
+	  if (ioctl(info->v4l_dev,VIDIOC_G_FMT,&fmt) < 0) {
+	    Error("ioctl(VIDIOC_G_FMT) Error");
+	  } else {
+	    fmt.fmt.pix.width=v4l_service->current_buffer->frame.size[0];
+	    fmt.fmt.pix.height=v4l_service->current_buffer->frame.size[1];
+	    fmt.fmt.pix.pixelformat = V4L2_PIX_FMT_GREY;
+	    if (ioctl(info->v4l_dev,VIDIOC_S_FMT,&fmt) < 0) {
+	      Error("ioctl(VIDIOC_S_FMT) Error");
+	    }
 	  }
 	}

@@ -255,7 +264,6 @@ V4lStopThread(camera_t* cam)
 void
 V4lThreadCheckParams(chain_t *v4l_service)
 {
-
   v4lthread_info_t *info;
   info=(v4lthread_info_t*)v4l_service->data;

@@ -266,23 +274,20 @@ V4lThreadCheckParams(chain_t *v4l_service)
     // STOPING THE PIPE MIGHT BE NECESSARY HERE

     // "start pipe"
-    if (ioctl (info->v4l_dev, VIDIOCGCAP, &info->vid_caps) == -1) {
-      perror ("ioctl (VIDIOCGCAP)");
+    if (ioctl (info->v4l_dev, VIDIOC_QUERYCAP, &info->vid_caps) == -1) {
+      perror ("ioctl (VIDIOC_QUERYCAP)");
     }
-    if (ioctl (info->v4l_dev, VIDIOCGPICT, &info->vid_pic)== -1) {
-      perror ("ioctl VIDIOCGPICT");
-    }
-    info->vid_pic.palette = VIDEO_PALETTE_RGB24;
-    if (ioctl (info->v4l_dev, VIDIOCSPICT, &info->vid_pic)== -1) {
-      perror ("ioctl VIDIOCSPICT");
-    }
-    if (ioctl (info->v4l_dev, VIDIOCGWIN, &info->vid_win)== -1) {
-      perror ("ioctl VIDIOCGWIN");
+
+    if (ioctl (info->v4l_dev, VIDIOC_G_FMT, &info->vid_fmt)== -1) {
+      perror ("ioctl VIDIOC_G_FMT");
     }
-    info->vid_win.width=v4l_service->current_buffer->frame.size[0];
-    info->vid_win.height=v4l_service->current_buffer->frame.size[1];
-    if (ioctl (info->v4l_dev, VIDIOCSWIN, &info->vid_win)== -1) {
-      perror ("ioctl VIDIOCSWIN");
+
+    info->vid_fmt.fmt.pix.width=v4l_service->current_buffer->frame.size[0];
+    info->vid_fmt.fmt.pix.height=v4l_service->current_buffer->frame.size[1];
+    info->vid_fmt.fmt.pix.pixelformat=V4L2_PIX_FMT_RGB24;
+
+    if (ioctl (info->v4l_dev, VIDIOC_S_FMT, &info->vid_fmt)== -1) {
+      perror ("ioctl VIDIOC_S_FMT");
     }
   }

diff --git a/src/thread_v4l.h b/src/thread_v4l.h
index 91710f7..0551e1d 100644
--- a/src/thread_v4l.h
+++ b/src/thread_v4l.h
@@ -30,9 +30,8 @@ typedef struct
   dc1394video_frame_t frame;
   //unsigned char *v4l_buffer;

-  struct video_capability vid_caps;
-  struct video_window vid_win;
-  struct video_picture vid_pic;
+  struct v4l2_capability vid_caps;
+  struct v4l2_format vid_fmt;

 } v4lthread_info_t;