Hi,
Am 13.09.2012 14:05, schrieb Hans de Goede:
> Hi,
>
> On 09/12/2012 04:36 PM, Frank Schäfer wrote:
>
> <snip>
>
>>
>> And a negative side effect is, that unknown pac7302 devices (with no
>> V4LCONTROL_ROTATED_90_JPEG entry in libv4lconvert) do not work.
>> With a consistent API behavior, they would work fine (output a rotated
>> image). Users would at least know that their device is working and most
>> of them know what to do next.
>> For image rotation, we still need to add an entry to libv4lconvert and
>> to modify it to invert the width and height values in v4l2_pix_format in
>> this case.
>
> That is a good point, unfortunately we are stuck with how we are doing
> things now, since changing things would break the kernel ABI.
Yeah, we can't break things. But I think this would be ABI fixing not
breaking. ;)
Actually...I'm not sure if this would be really an ABI change, because
the interface itself wouldn't change.
We would only fix a driver which passes a wrong value to userspace.
Its a question of interpretation...
>
> Also ...
>
>>
>> The device I have here is a good example: many people reported this
>> device as not working years ago, one of them even got a hint in a forum
>> that this could be a pac7302 device 2 years ago.
>> But with the gpsca-pac7302 driver, he got no picture and gave up.
>> And if I had not started q4vl2 from the terminal and had noticed the
>> error message from libv4lconvert, I would have needed much more time to
>> find out what's wrong...
>
> True, OTOH just having this fixed won't help a regular user, as he/she
> would still need to first add the new usb-id to the pac7302 driver...
Regular users, sure. But it would be a big step forward.
Adding new device IDs for testing is one of the easier things in the
Unix world.
Hans, I have a bunch of smaller things on my ToDo list which I want to
do first.
For now: maybe we can improve things by trusting the jpeg-header ?
That would mean removing the following section from
v4lconvert_decode_jpeg_tinyjpeg() :
if (data->control_flags & V4LCONTROL_ROTATED_90_JPEG) {
unsigned int tmp = width;
width = height;
height = tmp;
}
if (header_width != width || header_height != height) {
V4LCONVERT_ERR("unexpected width / height in JPEG header: "
"expected: %ux%u, header: %ux%u\n",
width, height, header_width, header_height);
errno = EIO;
return -1;
}
V4LCONTROL_ROTATED_90_JPEG is only used for the pac7302 devices and we
know that their header is correct.
And even in cases where the header is wrong, I think it would we better
to get a garbeled picture instead of -EIO.
Regards,
Frank
>
> Regards,
>
> Hans
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
