mark.paulus at verizonbusiness
Nov 14, 2007, 8:34 AM
Post #8 of 14
(8977 views)
Permalink
|
|
Re: [ivtv-users] [TEST] Please test my ivtv-i2c tree!
[In reply to]
|
|
Scott Reynolds wrote:
> On Nov 13, 2007, at 4:42 PM, Hans Verkuil wrote:
>
>> I heard more reports about compatibility problems and I traced it to
>> some recent videodev changes that break compatibility for kernels <
>> 2.6.19. So in order to test this you need a kernel >- 2.6.19.
>
> I looked into this last week and I'm pretty confident this is just a
> bogus version test. If you look at the lines near the errors
> mentioned, you'll see a check for kernel 2.6.13; however, struct class
> didn't gain .dev_attrs, and hadn't renamed .release to .dev_release,
> until 2.6.19.
>
> If you change both #if LINUX_VERSION_CODE lines from 13 to 19 you'll
> get it to compile (with a warning), but there will be another error
> v4l1-compat.c that I'm not entirely sure of how to handle. At least on
> my system it chokes on the __pure attribute to palette_to_pixelformat().
>
> The patch I used to get the latest v4l-dvb to compile is appended. It
> seems to work for me but I can't guarantee its correctness, YMMV, etc.
>
> --scott
>
> diff -r b9523a14ccea linux/drivers/media/video/v4l1-compat.c
> --- a/linux/drivers/media/video/v4l1-compat.c Sun Nov 04 14:34:05 2007
> -0200
> +++ b/linux/drivers/media/video/v4l1-compat.c Wed Nov 07 00:00:45 2007
> -0600
> @@ -145,7 +145,7 @@ const static unsigned int palette2pixelf
> [VIDEO_PALETTE_YUV422P] = V4L2_PIX_FMT_YUV422P,
> };
>
> -static unsigned int __pure
> +static unsigned int /* __pure */
> palette_to_pixelformat(unsigned int palette)
> {
> if (palette < ARRAY_SIZE(palette2pixelformat))
> diff -r b9523a14ccea linux/drivers/media/video/videodev.c
> --- a/linux/drivers/media/video/videodev.c Sun Nov 04 14:34:05 2007
> -0200
> +++ b/linux/drivers/media/video/videodev.c Tue Nov 06 19:31:57 2007
> -0600
> @@ -111,7 +111,7 @@ static void video_release(struct device
> vfd->release(vfd);
> }
>
> -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,13)
> +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,19)
> static struct device_attribute video_device_attrs[] = {
> __ATTR(name, S_IRUGO, show_name, NULL),
> __ATTR_NULL
> @@ -120,7 +120,7 @@ static struct device_attribute video_dev
>
> static struct class video_class = {
> .name = VIDEO_NAME,
> -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,13)
> +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19)
> .release = video_release,
> #else
> .dev_attrs = video_device_attrs,
>
I used this patch and managed to get the modules to compile and load
on my 2.6.18 debian/etch kernel. However, I have a question. Will
these modules now load in a deterministic manner? For instance,
on this box I have a PVR-150, a air2PC card, an Avermedia A180,
and a DVICO pcHDTV RT5 lite. My issue is that in myth, I need to
assign the proper /dev/device to the proper capture device.
To get around that with previous versions, I had some custom rules in
hotplug/udev that would create custom aliases, such as /dev/ivtv0,
/dev/bttv_vid0. However, this release has broken the mechanism that
created my /dev/ivtv0, so mythtv isn't happy.
Will all devices load and create /dev/video* in the same order all the time,
or is it still up in the air as to which device will create which
/dev/video*, based upon who's faster today????
This is a long outstanding issue with udev/hotplug, but someone needs to
fix it, if they can figure out how.
|
mark_paulus.vcf
ivtv-messages.txt
