Additionally, make sure a buffer gets enqueued again (even in error paths) after
it has been succesfully dequeued.
Tested-by: Dmitry Volyntsev <xeioexception@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
As of September 14 2012, v4l_enumstd() will return ENODATA
when a device's std field is set to 0. That is, the device
does not have a standard format. In order to properly
handle this case, v4l2_set_parameters should catch the
ENODATA code and break instead of failing.
Below is the v4l2-core commit describing this change.
>>commit a5338190efc7cfa8c99a6856342a77d21c9a05cf
>>Author: Hans Verkuil <hans.verkuil@cisco.com>
>>Date: Fri Sep 14 06:45:43 2012 -0300
>>
>> [media] v4l2-core: tvnorms may be 0 for a given input, handle that case
>>
>> Currently the core code looks at tvnorms to see whether ENUMSTD
>> or G_PARM should be enabled. This is not a good check for drivers
>> that support the STD API on one input and the DV Timings API on another.
>> In that case tvnorms may be 0.
>> Instead check whether s_std is present (for ENUMSTD) or whether g_std or
>> current_norm is present for g_parm.
>> Also, in the enumstd core function return ENODATA if tvnorms is 0,
>> because in that case the current input does not support the STD API
>> and ENUMSTD should return ENODATA for that.
>>
>> Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
>> Reviewed-by: Sakari Ailus <sakari.ailus@iki.fi>
>> Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Whenever av_gettime() is used to measure relative period of time,
av_gettime_relative() is prefered as it guarantee monotonic time
on supported platforms.
Since v4l2 is compiled only on Linux where av_gettime_relative()
is monotonic, a small wrapper function av_gettime_monotonic() is used
(should be inlined the compiler) to put emphasis on that fact.
Signed-off-by: Olivier Langlois <olivier@trillion01.com>
Reviewed-by: Nicolas George <george@nsup.org>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Setting the v4l2 device parameters may influence video properties
such as width, height and pixel format.
This patch moves v4l2_set_parameters to be executed before fetching
the video properties, avoiding errors such as (pal vs ntsc):
The v4l2 frame is 829440 bytes, but 691200 bytes are expected
Signed-off-by: Federico Simoncelli <fsimonce@redhat.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
With some (buggy) drivers, the VIDIOC_G_STD ioctl returns a std_id that cannot
be matched with any of the enumerated v4l2_standard structures (for example
std_id = 0 or std_id = 0xffffff). Do not fail when we reach the end of the
enumeration without a valid match.
Fixes ticket #2370
Note: This commit message has been modified by Giorgio Vazzana, the original
commit message was:
"Fixed regression for mandatory VIDIOC_ENUMSTD support by v4l2"
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
In particular, print the standard supported by the selected input.
Additionally, use PRIx64 everywhere when printing standards.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
* commit '7950e519bb094897f957b9a9531cc60ba46cbc91':
Disable deprecation warnings for cases where a replacement is available
Conflicts:
libavcodec/avpacket.c
libavcodec/pthread.c
libavcodec/utils.c
libavdevice/v4l2.c
libavfilter/avfiltergraph.c
libavfilter/buffersrc.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
Also disable it by default since it looks currently buggy. The
usefulness of such library is mostly limited to backward compatibility
with very old devices.
This is consistent with the rest of the file and makes listing of
emulated formats possible when ffmpeg is compiled with libv4l2.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
An input channel could have been previously set with another application, like
v4l2-ctl, so if no input channel is specified use the previosly selected one.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
The current code returned a period=0, resulting in identical pts for all frames
after time-filtering. This is because AV_TIME_BASE_Q={1, AV_TIME_BASE} and
not {AV_TIME_BASE, 1}. With this patch the correct period in microseconds is computed.
struct buff_data contains a pointer to struct video_data, so passing the
file descriptor again is redundant.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
In particular:
1) save errno before it (possibly) gets overwritten by other calls
2) do not forget to enqueue the buffer again in case of error
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This will avoid the possibility that we dequeue more buffers than we
have obtained from the v4l2 driver.
Fixes ticket #1570
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Generally speaking, there are two types of v4l2 devices [1]:
1) devices that support a standard, like PAL or NTFS (tv cards, for example). For
this class of devices the framerate is fixed by the standard (for example PAL uses
25 fps) and the v4l2 driver cannot usually negotiate a different framerate (unless
it can skip frames on the driver side, to save I/O bandwidth).
2) devices for which the notion of standard does not make sense (webcams, for example).
For these devices it is usually possibile to request a desidered framerate.
In either case, the desidered frame rate can be requested when the VIDIOC_G_PARM
ioctl returns the V4L2_CAP_TIMEPERFRAME flag in the capability field.
Currently the code does not check for V4L2_CAP_TIMEPERFRAME and supports only the
second category of devices, returning a time per frame of 0/0 for devices in the
first group that do not permit to negotiate the framerate.
This patch adds support to read the correct framerate in all cases.
[1] http://linuxtv.org/downloads/v4l-dvb-apis/standard.html
Signed-off-by: Stefano Sabatini <stefasab@gmail.com>
Since the user is expected to choose the standard by name (with -standard
option), add the possibility to list all the supported standards.
Signed-off-by: Stefano Sabatini <stefasab@gmail.com>
Because libavdevice/v4l2 is a demuxer, it makes sense to look for
decoders instead of encoders when listing the formats supported by the
device.
Signed-off-by: Stefano Sabatini <stefasab@gmail.com>
Giorgio Vazzana
Dmitry Volyntsev
Michael Niedermayer
Andre Wolokita
Gabriel Dume
Olivier Langlois
Lukasz Marek
Federico Simoncelli
Derek Buitenhuis
mrlika
Carl Eugen Hoyos
Rémi Denis-Courmont
Diego Biurrun
Clément Bœsch
Nicolas George
Anton Khirnov
Stefano Sabatini
Stephan Hilb