Hi Laurent Le jeudi 06 mars 2008 à 23:30 +0100, Laurent Pinchart a écrit :
Hi Thierry,
On Thursday 06 March 2008, Thierry Merle wrote:
Laurent Pinchart a écrit :
On Monday 03 March 2008, Thierry Merle wrote:
Jiri Slaby a écrit :
Here I would like to know if the the commands passing interface to the helper daemon introduced in this patch is OK, or alternatively propose some other idea ;).
I have committed your patch as is http://linuxtv.org/hg/~tmerle/v4l2_extension/ Now will begin a driver enhancement. I will do that on usbvision because I know it. The first step will be to extend the supported video formats (2 sub-steps: 1-just enable hardware pixel format capabilities in usbvision, 2-allow the helper daemon to extend usbvision pixel format capabilities).
I haven't followed v4l2_extension development closely, so I'm a bit puzzled by this. Are the modifications you made to the usbvision module for testing purpose only ? My understanding of v4l2_extension is that it should work completely transparently and must not require any change to v4l2 drivers.
In fact I meant remove from usbvision the software decompression algorithm in order to put it in the helper daemon. I will restrict the list of supported pixel formats to the ones that the hardware can output without software decompression.
That's the whole point of having userspace decompression. Those changes are perfectly ok.
I hope that the sole modification for a standard base driver will be the add of v4l2ext_register/v4l2ext_unregister calls.
I'm not very happy with that. Couldn't this be done automatically at the videodev level ? As "v4l2ext" is a temporary solution, I'd hate seeing all drivers being modified. This should be transparent to the driver itself, otherwise we will have to modify them back again when v4l2ext will be replaced by a real userspace library.
I will try to do that, in a second time; I do not see any blocking point, the v4l2ext_register can be done in video_register_device, for VFL_TYPE_GRABBER. Priority #1 is to uncompress a frame in userspace (I have a cruel lack of time these days). But rebemember that the base driver should be ioctl-v4l2 compliant, since each ioctl is hooked in order to call the helper daemon. There will be some effort to do for old webcam drivers.
Cheers,
Laurent Pinchart
I hope I will be able to post to the right list at least one time :) Cheers, Thierry
-
Thierry MERLE