| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
The -A option is unreliable for the vivid driver due to a known
CEC emulation bug. Drop the -A option for now since it causes
random warnings and failures that make it unusable for
regression testing.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
The vivid's CEC emulation isn't perfect (it doesn't handle
Arbitration Lost situations properly). So disable some warnings
when the vivid emulation is detected.
This should be reverted once vivid is eventually fixed.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Do some autodetection magic and validate VDEV and capture format
strings passed to the script.
It will now try to use as many GStreamer features as supported by
the current version, and stream on all supported capture formats,
if no capture format is specified at the command line.
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
|
|
Change the name of GStreamer (it was written with wrong cases);
Add notes about GStreamer and script issues.
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
|
|
There are some whitespaces used there instead of tabs.
No functional changes.
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
|
|
This script allow testing all supported capture formats from vim2m
with qvidcap.
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
|
|
This script allows testing the vim2m driver and the corresponding
GStreamer support.
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
|
|
Fixes commit 8e695c1d3403 ("v4l2-ctl-streaming.cpp: use cv4l-helpers.h")
which used g_bytesused() and g_data_offset() instead of explicitly
specifying the plane index as argument.
So it would only take bytesused and data_offset from the first plane.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Reported-by: Benoit Parrot <bparrot@ti.com>
|
|
If the link is dynamic, then it is a failure if MEDIA_LNK_FL_IMMUTABLE
is set (the two are mutually exclusive). But the test was the wrong
way around. Fix this.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Reported-by: Tomasz Figa <tfiga@chromium.org>
|
|
The previous commit accidentally broke one error code check where
EINVAL was changed to EBADR when it shouldn't.
Revert that part of the commit.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
If requests are used when they shouldn't, or not used when they should,
then EBADR (Invalid request descriptor) is now returned instead of EACCES.
Update the compliance tests accordingly.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Synchronize with the latest headers from the media master tree.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
My e-mail has changed over the years. Due to that, people still
send me bug reports to e-mails I don't use anymore for upstream
stuff, or even send e-mails to addresses that doesn't exist
anymore.
So, let's put an order at the house, using just the canonical
e-mail mchehab@kernel.org, as this e-mail should always point
to my current e-mail preference.
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
|
|
do_handle_cap() returns -2 if EPIPE what returned, and it should
also return -2 when stream_count reaches 0 since -2 indicates
end-of-stream.
But when capturing the error code was only checks for -1 instead
of < 0. Fix this as well.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
When streaming for m2m devices the stream-count should be exact
and work even for e.g. --stream-count=1.
This means that when filling in the initial output buffers, no more
than stream_count buffers should be filled, even for count values
less than the number of allocated buffers.
This also means that STOP should be issued right after filling in
the buffers if stream_count has already reached 0.
This patch also is more precise when > or < is printed and when
the buffer contents (--verbose) is shown.
'<' is shown after a capture buffer is dequeued.
'>' is shown after an output buffer is queued.
The buffer contents is always shown after a buffer is dequeued.
For capture or output devices the stream count refers to the number
of dequeued buffers, and the number of queued buffers is
more than that (number of allocated buffers + stream_count).
For m2m devices the stream_count refers to the number of queued
output buffers.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Running 'ir-keytable --device /dev/input/event24 -p rc5' does not work
since ir-keytable cannot find an rc device via the input device name.
Simply remove this options from ir-keytable since it is not useful and
is only misleading.
Signed-off-by: Sean Young <sean@mess.org>
|
|
dvbv5-zap.c: In function ‘main’:
dvbv5-zap.c:1055:16: warning: assignment discards ‘const’ qualifier from pointer target type [-Wdiscarded-qualifiers]
args.dvr_pipe = default_dvr_pipe;
^
Just drop const from default_dvr_pipe.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Synchronize with the latest headers from the media master tree.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
noinline_for_stack annotations were added to codec-fwht.c.
That's kernel specific and can be defined away.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
This test transmitted Image View On from LA 0xf. Which is fine if
this is a display that pulls down the HPD when in Standby, but it
fails for all other displays. In that case you just use the current
logical address.
Fix this by first trying to transmit with the current LA, then on
error transmit again, now from LA 0xf.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Signed-off-by: Sean Young <sean@mess.org>
|
|
dvb_fe_open_fname() takes ownership of fname if the function succeeds, but
also in two of the error paths (e.g. if the ioctl FE_GET_PROPERTY fails).
Adjust dvb_fe_open_fname() so it copies fname rather than taking ownership
(and passing that to params). This makes the code cleaner.
Signed-off-by: Sean Young <sean@mess.org>
|
|
Signed-off-by: Sean Young <sean@mess.org>
|
|
While testing the program with Valgrind, it would keep reporting
about non-freed data. The main reason is that, when called in
non-record and non-monitor mode, the while() loop doesn't check
for the timeout indication.
Solve that and be sure that all strdup() vars will be freed
too.
On my tests, on normal mode, on monitor mode and on record mode,
it now shows no memory leaks with Valgrind.
Having zero leaks in Valgrind is important mostly because we
don't want the libraries to leak. Having all memory allocated
internally at the program freed means that, if Valgrind will
report any memory leaks in the future, it would belong to
the library code.
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
|
|
Let's not let the arguments struct with some random value.
Signed-off-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
|
|
This fixes conversion of:
http://lirc.sourceforge.net/remotes/goldstar/VCR
Signed-off-by: Sean Young <sean@mess.org>
|
|
Signed-off-by: Sean Young <sean@mess.org>
|
|
set the in/out fmt variables in streaming_set_m2m
This is needed later to check for stateless
fwht pixel format.
Signed-off-by: Dafna Hirschfeld <dafna3@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
'read_write_padded_frame' should check that the
expected size to read/write is not larger than
the size of the buffer.
Signed-off-by: Dafna Hirschfeld <dafna3@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
In streaming_set_m2m, make sure to close all file pointers
and file descriptors before returning.
Signed-off-by: Dafna Hirschfeld <dafna3@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Added a few missing newlines and log when the en/decoder is stopped.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
The concise buffer info didn't include the timestamp flags,
which was confusing. So just show them.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
If after clearing the timestamp-related flags the value was
0, then the string would end with ", ".
That is ugly, so rework the code.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
This causes the following lircd.conf to be converted incorrectly.
http://lirc.sourceforge.net/remotes/cambridge_audio/X40A
Signed-off-by: Sean Young <sean@mess.org>
|
|
print_concise_buffer() shows buffer information, but not whether it
was a capture or output buffer. This is confusing when dealing with
m2m devices, so add a 'cap' or 'out' prefix.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Don't show crop info when using --all, only show the selection info
(crop doesn't add any useful information).
Also show the buffer type for which the selection info is shown.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Add support for 16 bit Bayer formats:
-V4L2_PIX_FMT_SBGGR16
-V4L2_PIX_FMT_SGBRG16
-V4L2_PIX_FMT_SGRBG16
-V4L2_PIX_FMT_SRGGB16
Tested using vivid included in linux v5.0-rc8.
Signed-off-by: Daniel Gomez <daniel@qtec.com>
Signed-off-by: Ricardo Ribalda <ricardo@ribalda.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Add support for 10 bit Bayer formats:
-V4L2_PIX_FMT_SBGGR10
-V4L2_PIX_FMT_SGBRG10
-V4L2_PIX_FMT_SRGGB10
Previous BAYER10 format declared (V4L2_PIX_FMT_SGRBG10) now is grouped
with the new list without the need of tmp buffer.
Update v4lconvert_10to8 function:
- Renaming function name to keep naming convention with the
other bayer10p conversion function:
v4lconvert_10to8 -> v4lconvert_bayer10_to_bayer8
Tested using vivid included in linux v5.0-rc8.
Signed-off-by: Daniel Gomez <daniel@qtec.com>
Signed-off-by: Ricardo Ribalda <ricardo@ribalda.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
make -> makes
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Fix lots of places where 'than' instead of 'then' should be used.
Reported-by: Jason H <jhihn@gmx.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Signed-off-by: Sean Young <sean@mess.org>
|
|
sizeof(name) is 64 for media_v2_entity but 32 for
struct media_entity_desc. So only consider the first 32 bytes
(31 characters) for the duplicate name tests and matching what
ENUM_ENTITIES reports with what G_TOPOLOGY reports.
Also relax the check of the possible interface types to allow
alsa as this is upcoming with the Media Device Allocator API.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
|
|
There were compliance failures when testing vimc due to a wrong setup
for Raw Capture 1. That is now fixed.
Tests have also been added were the subdevs are unbound before the
main vimc driver.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Signed-off-by: Sean Young <sean@mess.org>
|
|
The vimc driver consists of multiple modules, but rmmod only
unloaded the vimc module. Add all vimc modules to the rmmod line.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Use random sleep values so device nodes are closed in a different
order for each test.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Just use sleep directly.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
When combining -z with --list-devices all devices belonging
to the given platform device are shown, except for the media
device itself. Show the media device as well.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
This helps test the case where you unbind the device while
the device nodes are still open.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
|
Using '-z media-bus-info --list-devices' will show all devices
as specified in the media topology of the media device with the
given bus info.
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
|
