Add 'all' option to the following test tools to use
V4L2_FMTDESC_FLAG_ENUM_ALL if supported by the driver:
- vidcap --list-formats(-ext) options
- meta --list-formats-meta option
- vidout -list-formats-out(-ext) options

Update functions prototype to provide an additional parameter.

Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>

If V4L2_FMTDESC_FLAG_ENUM_ALL flag is supported, test if all
pixel formats list with VIDIOC_ENUM_FMT without the flag been set
is a subset of the list created with the flag.
Also Test that the flag is cleared of calling VIDIOC_ENUM_FMT.

Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>

The 'cv4l_queue q' variable went out of scope too early,
causing a core dump when g_dataptr was called later.

I suspect it was pure luck that this ever worked, but the changes
to support more than 32 buffers in combination with vivid supporting
this new feature, caused this bug to surface.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Signed-off-by: Gregor Jasny <gjasny@googlemail.com>

musl since version 1.2.0 uses 64-bit time_t and suseconds_t on all
platforms, even 32-bit. Cast to avoid warnings and potentially future
proof for when glibc and others gain support.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Use int for this field instead of time_t since it is at most
about 100*3600 seconds.

It also avoids confusion about the size of time_t, since that can
be 32 or 64 bits.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Under musl, if a format string has an integer followed by %s, a mismatch
between types can cause the second half of the integer to be interpreted
by %s.

Eg: printf("%d %s", 64bittype, string);

will crash, especially on 32-bit big endian.

The reason these are cast to __u64 is because time_t and suseconds_t
are 64-bit under musl, even on 32-bit platforms. __u64 helps avoid
any truncation issues that may or may not arise.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
[hverkuil: the change to cec-follower was dropped]

Found with: modernize-use-auto

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Found with: modernize-loop-convert

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
[hverkuil: dropped the changes to the generated (re)trace-gen.cpp files]

musl uses 64-bit off_t by default on all platforms. Cast and fix format
to work everywhere.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Unlike libc, kernel headers use long long for 64-bit types.
__SANE_USERSPACE_TYPES__ needs to be defined for the exceptions.

Fixes -Wformat warnings about llu being used instead of lu.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Support for this buffer capability was missing. Add it.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Add missing file.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Sync up with the latest media_stage master tree.

This also drops include/linux/compiler.h since it is no longer
needed (and hasn't been for a long time).

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Now that vivid is able to virtually connect HDMI inputs
to HDMI outputs, that also means that by default the CEC
devices no longer are connected, and you need to set up that
loopback configuration manually in test-media.

Add the necessary commands for that.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Make the code easier to understand and improve the comments.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Reported-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>

The tests to verify correct VIDIOC_CREATE_BUFS behavior were
insufficient, causing drivers with bugs in their implementation
to pass.

Add new tests and improve existing tests so such corner cases are
now caught.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

If matchFormats() fails for multiplanar video formats, then it
is helpful to log what the differences were.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

v4l_format_s_sizeimage() can now be used with any buffer type, except
for the raw VBI buffer types, since those do not have a 'sizeimage'
equivalent.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Add a -W/--which argument to media-ctl to select which state to operate
on. Default to the ACTIVE state to preserve the current behaviour.

Despite the fact that all values set on the TRY state are lost when
media-ctl terminates, support for the TRY state is useful in order to
retrieve the default configuration of subdevs, or to try configurations.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>

Pass a 'which' argument to all functions to select which state to
access. Hardcode the value to V4L2_SUBDEV_FORMAT_ACTIVE in a single
location in main(), to preserve the existing behaviour. This prepares
media-ctl for support of the TRY state.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>

Many functions exposed by libv4l2subdev have been added over time
hardcoding usage of the V4L2_SUBDEV_FORMAT_ACTIVE state. Extend them
with a 'which' argument to allow callers to access the TRY state as
well. Update existing callers to pass V4L2_SUBDEV_FORMAT_ACTIVE
unconditionally, preserving the existing behaviour.

To support accessing the TRY state when getting or setting frame
intervals, set the V4L2_SUBDEV_CLIENT_CAP_INTERVAL_USES_WHICH client
capability.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>

v4l2-test-time32-64.cpp included compiler.h, which checks
_LIBCPP_VERSION. This only works against libc++ when a C++ header has
already been included, which wasn't the case here.

The <version> header is the C++20 method of defining _LIBCPP_VERSION,
but for older versions, <ciso646> works as an alternative, so include
that in compiler.h.

compiler.h is for C as well as C++ though, so use __cplusplus to check
for a C++ compiler before including <ciso646>.

Signed-off-by: James Le Cuirot <chewi@gentoo.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Remove --test-keymap since the command line processing is broken if
--test-keymap is specified with other arguments. A keymap can simply
be tested by loading it.

Also ensure `--auto-load` properly excludes other options.

Signed-off-by: Sean Young <sean@mess.org>

Some ARIB strings in operation requires that
codepoints from the 'extra symbols' charset fallback to 'kanji'.
Remove the range check for 'extra symbols' to accept those strings.

Signed-off-by: Akihiro Tsukada <tskd08@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>

Change/Add entries according to the ARIB STD B24 table 6.19.

Signed-off-by: Akihiro Tsukada <tskd08@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>

Sometimes the modprobe vivid fails, but because dmesg is set to
the 'notice' level it is not clear why it fails since the usual
kernel info messages are filtered out.

Switch to dmesg -n info before each modprobe so we can actually
see what is going on.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

The vivid driver does not produce reliable sequence counters,
the first sequence counter may be > 0, and there are gaps in the
counters when there shouldn't be.

Until this is fixed (work in progress) replace the warning when
an unexpected vivid_ro_ctrl value is seen by an info message.

Once vivid is fixed this can be reworked.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

The v4l2-dv-timings.h header was never copied from the kernel
source to the local include/linux directory. Add it.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

The metadata API supports a new flag, V4L2_FMT_FLAG_META_LINE_BASED. It
can be set for metadata formats only, and is the only flag that can be
set for metadata formats. When set, the width, height and bytesperline
fields must not be zero. Add corresponding tests.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Set len_routes of struct v4l2_subdev_routing. ENOSPC error code is no
longer used, i.e. having room for fewer routes than exist in the
configuration is not considered an error anymore.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Set len_routes of struct v4l2_subdev_routing. ENOSPC error code is no
longer used, i.e. having room for fewer routes than exist in the
configuration is not considered an error anymore.

Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Detect if this flag is set and report it.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Update v4l-utils to the latest kernel headers.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

These meta formats are not yet enabled (under #ifdef __KERNEL__),
so drop these for now.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

It can be handy to see what the current input or output name is of
a given v4l2 device: that often gives useful information to determine
which device to use. So if --verbose is added to --list-devices, then
it will report that information in parenthesis after the device name.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

If the --list-devices option is used, then just exit after listing all
devices. Do not continue and report the driver info of /dev/cec0, that's
unexpected.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Transmitting pulse_length is broken and never worked. The most common
user of this protocol is Sony, which already has its own encoder.

Signed-off-by: Sean Young <sean@mess.org>

The optional header was not included when transmitting.

Signed-off-by: Sean Young <sean@mess.org>