We need to unload media before we delete the player.
Task-number: QTBUG-37447
Change-Id: I09263214c425339c0d8c192091aeaa6e3fb5331f
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
Instead of setting the volume on the audio session, which is shared by
all QMediaPlayers, we now set the volume on the media player's own audio
stream. This results in all QMediaPlayers correctly having independent
volumes.
[ChangeLog][QtMultimedia][Windows] QMediaPlayer::setVolume() does not
affect the volume of other QMediaPlayers anymore.
Task-number: QTBUG-30317
Change-Id: I8ea8ec47fc86127da01dc5c8247fb6f72c834630
Reviewed-by: Wouter Huysentruit <wouter_huysentruit@hotmail.com>
Reviewed-by: Andy Nichols <andy.nichols@digia.com>
This fixes the issue that when we use QSoundEffect we dont need to get
permission to use the microphone. This was because we were defaulting
to PlayAndRecord. Now we only switch to the PlayAndRecord category when
we try and use an input device.
[ChangeLog][QtMultimedia][iOS] Using QSoundEffect (or SoundEffect in
QML) no longer requires permission to use the microphone.
Change-Id: I94535215497a718005c280bfd84b428b4bf1b16a
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
Clear the tag list before syncing with reference set so old values
are removed and explicitly remove invalid QVariants from the reference
set so that the type correction doesn't result in invalid values being
written instead.
Change-Id: I7d1bdc95cd2a2d601720db84c9b3ef629477bc99
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
The old textures won't match anyways. So there is no need to keep them.
Change-Id: Id3482333d10cf022d04076ec0f5c7df475c522ae
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
Reviewed-by: Andy Nichols <andy.nichols@digia.com>
All formats are now tested.
Also, don't use the alpha channel. This is not quite correct for formats
with alpha channel but for video probably a good idea.
Change-Id: I4a2404d7c96023bb1e3fca41e1f224d23347f290
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
Reviewed-by: Andy Nichols <andy.nichols@digia.com>
This way it also works custom mkspecs
Change-Id: I7414aac775b797da3487e9625c7212486118b532
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
Reviewed-by: Andy Nichols <andy.nichols@digia.com>
The reference counting was not being done for AVCaptureSession so it was
being destroyed before we got a chance to properly clean it up (which
lead to crashes). We also make sure to remove any observers from
AVCaptureSession now before destroying it.
Task-number: QTBUG-37109
Task-number: QTBUG-29955
Change-Id: Ia9b49ad1eab01b4f7424e2a1c699d903cd9bf902
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
Before we were reusing the active surface when we switched videos. This
leads to an inconsitency in states, and the sourceRect property of the
video surface does not get updated.
Task-number: QTBUG-28655
Change-Id: Ie29bf1d9b1c11b6f51a869253c730202001c07cf
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
When the code that supports iOS was added, we introduced a bug where QML
videos would be played y-inverted on OS X. This is because we made no
effort to y-invert the Framebuffer Object before rendering the texture
in the SceneGraph. Now we render the video the the FBO y-inverted, so
there is no need to y-invert the resulting texture.
Task-number: QTBUG-35955
Change-Id: I41af1aaae57923b9972b5be5ec65f7d2a97d77c5
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
Some debug code made it into the released code, so each time you used
the QWidget based video player the terminal would print data about the
window frame size.
Change-Id: I86eb00ce5edb23b7a2abf6a63893cd17aaeb0ee3
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
There is an assumption in QtMultimedia that the first audio device
returned by QAudioDeviceInfo::availableDevices is the default device, so
we must make an effor to make sure this is true. This commit should fix
the issue on OS X.
Task-number: QTBUG-36638
Change-Id: Id388d7218b465cb29d826f46ee825e982c5f7ffc
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
QAudioOutput::setVolume stopped working for CoreAudio when it was ported
to live in it's own plugin. This was because it was not possible to set
the volume of QAudioOutput in iOS. Now the functionality has been
restored and added for iOS as well. For OS X we use the old method of
setting the volume property of the AudioUnit. On iOS it is not possible
to set the volume on a per AudioUnit basis, so we now manually modify
the buffer contents (the same we do for QAudioInput already).
Task-number: QTBUG-36756
Change-Id: I42b5892fe5534217043fa55e7b5b9a4ce824050d
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
The available sample rates for a given device are the nominal sample
rates reported by the device. It is possible to use other sample rates
and CoreAudio will take care of the conversion. So what we will do is
report what rates are available for a device, but not explicitly require
those sample rates to have a valid format.
Task-number: QTBUG-36265
Change-Id: Idbbdeacbb6bc1fe434bcd8dec519ad70d4ccd545
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
VideoOutput will now take into account the camera sensor orientation to
adjust the viewport orientation, we don't need to apply a rotation on
the Android Camera anymore.
Change-Id: Ia7639f0a5711ab6cc6f80b9716bc1a6f389499b4
Reviewed-by: Christian Stromme <christian.stromme@digia.com>
The class allows to get the list of available cameras on the system as
well as getting some static information about them such as their unique
ID, display name, physical position and sensor orientation.
This makes QCamera::availableDevices() and QCamera::deviceDescription()
obsolete.
This patch contains the API, documentation and auto-tests but not the
actual implementation by each backend (except for retrieving the default
camera device).
[ChangeLog][QtMultimedia] Added new QCameraInfo class
[ChangeLog][QtMultimedia] QCamera: availableDevices() and
deviceDescription() are deprecated, use QCameraInfo instead
Change-Id: I64fd65729ab26a789468979ed5444ee90bb82cd0
Reviewed-by: Christian Stromme <christian.stromme@digia.com>
Pass the top-level window's window group to mm-renderer,
as required by mmr.
Change-Id: I2a2e8b4aa48f5c2292b03593c6d528068f383b5c
Reviewed-by: Bernd Weimer <bweimer@blackberry.com>
Reviewed-by: Rafael Roquetto <rafael.roquetto@kdab.com>
Reviewed-by: Fabian Bumberger <fbumberger@rim.com>
When -no-opengl was used then it was not automatically pulling in the
needed Windows libraries as it was not loading opengl.prf to get them
from.
Change-Id: I9b73e5ee01da964a64c71b6261ff4ed98ce6f1a9
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
On imx6 there is a platform-specific API to implement
memory-mapping from gstreamer-buffer to opengl-texture.
This plugin uses this API to avoid CPU-based-memory-copy.
This allows fluid video-playback up to 1080p with very little CPU load.
Before even 720p used one cpu-core completely and wasn't fluid.
Change-Id: I0c33eb5d475393a65459291ce9290fa0753de4a5
Reviewed-by: Andy Nichols <andy.nichols@digia.com>
This enables writing gps positioning information to captured images.
Change-Id: I663f7b94b5e6040865358df1ae45d67b73c54f92
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
GStreamer tags are typed and the correct type must be used when inserting
a value into a GstTagList or subsequent merges or data accesses can crash
because of invalid casts.
Found while adding additional mappings for GPS values.
Change-Id: I95ab40a480a4685bf4e69064315557faa9de288e
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
Commit 773dd01 in qtbase introduced a general mingw platform scope, which
is cleaner and more flexible than matching the spec name.
(see 278152fff for a similar commit in qtbase).
Change-Id: Ib46b9e83f690b0b26933b73764c7d7d4b031d792
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
Trying to heuristically guess this kind of thing is almost always a bad idea.
Expect the creator of the output to explicitly tell us the category instead
of guessing.
This fixes QAudioOutput not respecting system volume on Sailfish.
Change-Id: If1d05192f513eb54fdfbd1df217286f329b2bfe8
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
The camera itself can have a mode set that causes the image to be flipped
so instead of always mirroring the image that is taken from the camera we
check for the supported modes first and then check the mode and only
mirror if it is set. Otherwise we assume that it does not need
horizontally flipping but that it needs vertically flipping which seems to
be the standard for cameras on Windows.
[ChangeLog][QtMultimedia][Windows] Fixed the incorrect mirroring of the
image from the camera
Task-number: QTBUG-30365
Change-Id: I166b1f354e8d91c9a6c64f64164d782b52df98d8
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
The number of bytes per line was missing.
Change-Id: I0afbdcfd6d7195b7beb3fd09f4ed262f756be848
Reviewed-by: Denis Kormalev <dkormalev@ics.com>
Reviewed-by: Christian Stromme <christian.stromme@digia.com>
Don't try to apply video recording settings until a camera is actually
started.
Change-Id: I5348b91669835b9d776f56a3f49e3f8a11533d15
Reviewed-by: Christian Stromme <christian.stromme@digia.com>
Fixes the "libqavfmediaplayer.dylib 0x0000000110fa7c8c -[AVFMediaPlayerSessionObserver unloadMedia]" crash on Macs.
The problem was: writing to memory that had already been released. If not sure, one should always check if the objects exist before deleting it.
Solution tested on OSX 10.7, 10.8, 10.9
[ChangeLog][qtmultimedia][avfmediaplayersession]
Task-number: QTBUG-34213
Change-Id: Iac108711851c348e96e73542b4e71653007eeb54
Reviewed-by: Andy Nichols <andy.nichols@digia.com>
It seems all Android media players share the same video buffers, which
results in textures containing frames for the wrong media player.
When getting a new frame, we now copy it into a FBO in order to avoid
that another media player overwrites the frame being shown on screen.
Task-number: QTBUG-35868
Change-Id: I6701cf7368a3ef9e73d649c3ece1f206cafd5bb3
Reviewed-by: Christian Stromme <christian.stromme@digia.com>
The custom focus point passed to the camera was in viewport orientation
but should be in sensor orientation. The two might differ.
Task-number: QTBUG-36208
Change-Id: Id272402090c2814e02abc527c0f6a6e932a60081
Reviewed-by: Christian Stromme <christian.stromme@digia.com>
Now based on fixed-point arithmetic and uses pre-generated lookup tables
for YUV coefficients and RGB clamping.
The new implementation is on average 2x-3x faster than the previous one
(tested on the Samsung Galaxy S4).
Change-Id: I1daf12f7f9b2f2334e90e0ede79e6d83800f1db6
Reviewed-by: Denis Kormalev <dkormalev@ics.com>
Reviewed-by: Christian Stromme <christian.stromme@digia.com>
Don't set profiles if no settings are specified.
Apply all settings before starting a pipeline as the mode can switch
without being restarted and incompatible video recording settings can
prevent the pipeline starting even in image capture mode.
Set audio encoding settings and encoder profiles if they are supplied.
Change-Id: I06febf977c2cae306383f9dbaae0f81f531b4757
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
Orientation tags need to be transformed from the string tag returned
by gstreamer to the orientation in degrees. Date tags need to be
inserted with gst_date_time_new_local_time. Finally setting a tag
value shouldn't clear all other tags.
Change-Id: I28922148251084c12cf6c93d9b097fa5df41da9d
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
Reloading the capture pipeline before recording is unnecessary and
introduces a big pause. Don't do it.
https://bugzilla.gnome.org/show_bug.cgi?id=649832 was resolved in gstreamer 0.10.23
Change-Id: I0de02af52dcc193bce7a4e3e61407ae1c2d3818c
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
Guessing badly prevents the camerabin pipeline from loading at all and
the fallbacks used are not representive of the capabilities of most
cameras. So either stay out of the process and let gstreamer negotiate
a resolution if it can, or use a resolution supplied through the viewfinder
settings control by someone hopefully better informed.
Task-number: QTBUG-30842
Change-Id: Iec2dcc2476f38822f9e0d02301e46a1f49b7c6da
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
- Quit the worker thread only when the worker is deleted. This makes
sure all events are processed before terminating the worker.
- Correctly cache the preview size even when it's an invalid one.
Task-number: QTBUG-36204
Change-Id: I76055984e8ece3f7f40dba7dd89d28a4faa1e72e
Reviewed-by: Denis Kormalev <dkormalev@ics.com>
Reviewed-by: Christian Stromme <christian.stromme@digia.com>
