Signed-off-by: André Almeida <andrealmeid@collabora.com>

Use v4l2-ctl to configure vimc output device before testing it.

Signed-off-by: André Almeida <andrealmeid@collabora.com>

v4l2-compliance tests deals with memory-to-memory devices (m2m) like
they are a single device, through the struct node. This is indeed the
most usual case. However, if we deal with m2m devices like they are two
different devices for the capture and for the output queue, this would
make things easier to test complex topologies where the capture and the
output queue are in different devices.
Split the node on testNode and related functions in two arguments:
struct node node and struct node node_m2m_cap. For now, use the same node in
both arguments.

Signed-off-by: André Almeida <andrealmeid@collabora.com>

The stream_count variable was decremented twice for regular
m2m devices like vim2m. Add ignore_count_skip to do_handle_out
and do_setup_out_buffers to avoid this.

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

If an m2m driver timeouts, then a message with newline was only
printed if 'stopped' was false. But if it was true, then there
still should have been a newline printed. Do this.

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

The input-event-codes.h now contains lines like:

	#define KEY_ZOOM                KEY_FULL_SCREEN

Ensure these parsed correctly.

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

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

Add support for the following pixelformat variants:

       V4L2_PIX_FMT_RGBX32
       V4L2_PIX_FMT_RGBA32
       V4L2_PIX_FMT_BGRX32
       V4L2_PIX_FMT_BGRA32
       V4L2_PIX_FMT_XBGR444
       V4L2_PIX_FMT_ABGR444
       V4L2_PIX_FMT_RGBX444
       V4L2_PIX_FMT_RGBA444
       V4L2_PIX_FMT_BGRX444
       V4L2_PIX_FMT_BGRA444
       V4L2_PIX_FMT_RGBX555
       V4L2_PIX_FMT_RGBA555
       V4L2_PIX_FMT_XBGR555
       V4L2_PIX_FMT_ABGR555
       V4L2_PIX_FMT_BGRX555
       V4L2_PIX_FMT_BGRA555

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

Update to latest media_tree master.

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

Currently ir-ctl just says the scancode is invalid, which is not
very clear.

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

If the driver under test does not support VIDIOC_PREPARE_BUF, then
the USERPTR streaming test will fail since length of the buffer was
zeroed. But that should only be done if PREPARE_BUF was supported,
so move the code that zeroes length, bytesused and userptr under
the 'if (ret == 0)' (the VIDIOC_PREPARE_BUF return code).

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

When showing the v4l2_frmsizeenum contents the min/max values weren't
logged for type V4L2_FRMSIZE_TYPE_CONTINUOUS. Add this.

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

Some displays only accept Standby from the active source,
so transmit the Active Source message before the Standby
message in the compliance test and the powercycle stress
tests.

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

The tests whether the expected devices are present post
a message if that's not the case and then exit. But that skips
the "Final Summary" line that is output at the end. Add that
in order for test scripts to see this.

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

If no cec device was found, bail out with a proper error
message.

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

If a modprobe foo doesn't actually load the module, then
report an error instead of mindlessly continuing.

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

Sync with the latest media_tree/master branch.

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

The compliance tests failed for output type V4L2_OUTPUT_TYPE_ANALOGVGAOVERLAY.
Fixed the tests.

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

If S_FMT(G_FMT) returns a different overlay format, then show
the differences.

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

Just getting the new format after a source change event isn't
enough in the case that -v option was specified by the user.

Instead let capture_setup optionally return the new format after
any '-v' changes are applied.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Reported-by: Dafna Hirschfeld <dafna3@gmail.com>

The power cycle stress test sent spurious Image View On messages.
While not harmful, it is not supposed to happen.

Add warnings when trying to run the power cycle tests for a TV
(makes no sense) or when trying to transmit messages when the
adapter is unconfigured.

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

One scancode can only be mapped to one keycode. In addition, the
toml is invalid so the keymap cannot be loaded.

Cc: Bastien Nocera <hadess@hadess.net>
Signed-off-by: Sean Young <sean@mess.org>

Some keymaps contain mappings in comments which should have not been
included.

Cc: Bastien Nocera <hadess@hadess.net>
Signed-off-by: Sean Young <sean@mess.org>

This changed in kernel v5.3. In earlier versions, EINVAL
will be returned.

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

The 'field' value of a buffer is checked against cur_fmt, but for
m2m devices there are two different formats: one for capture and
one for output. So for m2m capture devices check against cur_m2m_fmt
instead.

During streaming another issue occurs when it checks the 'field' of
the capture buffer against the field of the original output buffer,
but these can differ for e.g. m2m deinterlacers. So relax that test.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>

Add the new --test-power-cycle option: this will test how well a
display supports the Standby and Image View On messages. This is
the first version, and it is likely to be improved going forward.

Also improved the --stress-test-power-cycle option: it is now
possible to provide a max-sleep suboption to have the test sleep
for a random time from 0 to max-sleep seconds before transmitting
Standby or Image View On.

This will randomize the timings.

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

For compressed formats also check V4L2_BUF_FLAG_TIMECODE
for m2m devices. Also show the output buffer flags when
--verbose was specified.

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

The test for the stateful decoder that checks for a SOURCE_CHANGE
event when you start streaming needs to wait for up to one second
before calling dqevent. Otherwise the HW may not have had the
chance yet to raise this event.

So use select() to wait for an exception with a 1s timeout.

Reported-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>

Use 1U << 31 instead of 1 << 31 to avoid the 'shifting a signed
32-bit value by 31 bits problem' where the result is compiler dependent.

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

If a driver doesn't implement S_SELECTION, then these tests return
ENOTTY instead of EINVAL. So just checking against EINVAL isn't
good enough.

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

The fwht decoder part was missing the copy_cap_to_ref(): this copies
the decoded frame back to the internal reference frame. This used to
be done in the fwht decoder itself in the past, but after adding
stateless decoder support to vicodec this step was split off into
a separate function.

Also update the state structure initialization: it didn't set
ref_stride and was missing support for 1 or 4 components.

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

Add vicodec tests.

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

Add tests for stateful decoders.

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

1) No buffers are queued on either vb2_queue, but STREAMON is called
   for both queues. The next capture buffer queued should return empty
   with FLAG_LAST set.

2) The CAPTURE queue contains buffers, the OUTPUT queue does not. Now
   ENC_CMD_STOP. Check that the dequeued capture buffer is empty and
   has FLAG_LAST set.

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

The tests for the VIDIOC(_TRY)_EN/DECODER_CMD ioctls were not quite
in accordance to the spec. Update the tests.

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

Add tests to check that after setting the output format
the capture format is updated and that the crop rectangle
is set correctly.

Also warn if the output supports composition and fail if
the capture has crop or composition support.

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

Several functions had a 'from' argument to do operations from
that buffer index onwards, but it wasn't always correctly used.

That's now fixed.

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

make_devname() didn't call mi_media_get_device() to translate
a major/minor tuple to a human-readable device name.

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