Re: partial revert of "uvcvideo: set error_idx properly"

[Hans de Goede <hdegoede@redhat.com>]

<modified the CC list to be more appropriate>

Hi,

On 12/25/2012 05:56 AM, Mauro Carvalho Chehab wrote:

> The pwc driver can currently return -ENOENT at VIDIOC_S_FMT ioctl. This
> doesn't seem right. Instead, it should be getting the closest format to
> the requested one and return 0, passing the selected format back to
> userspace, just like the other drivers do. I'm c/c Hans de Goede for him
> to take a look on it.

I've been looking into this today, and the ENOENT gets returned by
pwc_set_video_mode and through that by:
1) Device init
2) VIDIOC_STREAMON
3) VIDIOC_S_PARM
4) VIDIOC_S_FMT

But only when the requested width + height + pixelformat is an
unsupported combination, and it being a supported combination
already gets enforced by a call to pwc_get_size in
pwc_vidioc_try_fmt, which also gets called from pwc_s_fmt_vid_cap
before it does anything else.

So the ENOENT can only happen on some internal driver error,
I'm open for suggestions for a better error code to return in
this case.

What I did notice is that pwc_vidioc_try_fmt returns EINVAL when
an unsupported pixelformat is requested. IIRC we agreed that the
correct behavior in this case is to instead just change the
pixelformat to a default format, so I'll write a patch fixing
this.

Regards,

Hans