The test doesn't work on OS X 10.9 and 10.10. It also doesn't work
on OS X 10.8, but the current CI system doesn't actually run tests
on 10.8. In the (future) event of us running tests on 10.8, this
change is required.
Change-Id: I6cbe6249aee2460353322d6f1254858238af3711
Reviewed-by: Yoann Lopes <yoann.lopes@theqtcompany.com>
We try to link in Qt5PlatformSupport, which depends on libqtfreetype, which in turn
is not installed. However we don't really need Qt5PlatformSupport, only QtCore private
headers for the JNI bridge.
Change-Id: Ic6d577081327eda7b80ce03c6fb7fd5cf024d05f
Reviewed-by: Yoann Lopes <yoann.lopes@theqtcompany.com>
The overload was added for NokiaX86 and RVCT and is bound for
removal.
Task-number: QTBUG-47260
Change-Id: I8eb0333e0e20323ab20f361186d9a47fd3ab9f24
Reviewed-by: Yoann Lopes <yoann.lopes@theqtcompany.com>
Implement QImageEncoderControl for camera still image capture.
It provides the functions to set camera capture resolution and
get supported resolutions list.
Task-number: QTBUG-46456
Change-Id: Ideb1aa02d420be3a30d588bebf31714fa4fa6415
Reviewed-by: Oliver Wolff <oliver.wolff@theqtcompany.com>
The header should not depend on indirect includes.
It uses QSize in inline functions, so include the
header.
Change-Id: I3a5fca458ded47c98b7f71749ad75660c6e049e9
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Using qFuzzyCompare in operator== makes it impossible to create a
consistent qHash() overload for the class later on (qFuzzyCompare
is not transitive).
It is unlikely that there will be numerical instability for frame
rates. If there is, qFuzzyCompare should be overloaded for either
this class or the FrameRateRange class.
Change-Id: Id975ea410b373e2987da160504f7618a8c498b5c
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
...and inline op!=. Mark them as nothrow.
More idiomatic C++ (symmetry between lhs and rhs).
Change-Id: I65ecbef961383897e4e17325ad62d45e1772fbb0
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Instead of an alias for QPair<qreal, qreal>.
Task-number: QTBUG-46563
Change-Id: I7e1ac68242810f7e5f7e161571a11f5de7850e29
Reviewed-by: Christian Stromme <christian.stromme@theqtcompany.com>
We were always using shared OpenGL contexts to render media player
frames into an OpenGL texture. There's no need to do that when there
already is a current context on the current thread. This happens in
non-QtQuick cases, when the OpenGL thread is also the main thread.
Change-Id: Icb97ed49609c764263007a43b6bb481e23768111
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
Also, use the correct shader for RGB formats, we should be using the one
that does swizzling in that case.
Change-Id: I77b6327501f833fb2f7b30411fbe853e975a59c7
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
Right now we will always use GLTextureHandle frames with Quick. This is
great in many cases, but not always. Applications that wish to examine
the frames (e.g. via video filters) will sometimes prefer frames in
system memory, even if this is slower to display.
Add QT_QUICK_NO_TEXTURE_VIDEOFRAMES which can be used to disable texture-based
video frames.
[ChangeLog] The environment variable QT_QUICK_NO_TEXTURE_VIDEOFRAMES
can now be used to disable OpenGL texture based video frames. This can be useful
in applications that wish to filter and process the video frames and are not
GPU based.
Change-Id: I5ca6f6d485d5bc6c2da8d47db563cd910c238ac9
Reviewed-by: Yoann Lopes <yoann.lopes@theqtcompany.com>
Lumia 530 has the same QTBUG-44838 device bug that gives black frames
when blitting to the target texture for image preview. Add it to
workaround blacklist.
Task-number: QTBUG-45920
Change-Id: Ia411793b90e487fa57b943ed7f01e370bfb84a1f
Reviewed-by: Andrew Knight <andrew.knight@intopalo.com>
We need to restart the video surface when the frame size changes.
Change-Id: I81af3cb40fb40f7d157174ac96d42213880fbacd
Reviewed-by: Timur Pocheptsov <Timur.Pocheptsov@digia.com>
The assert was always triggered, at least on OSX.
Change-Id: I9aabb6103b7f5540cff42facc448ffcf6a6de511
Reviewed-by: Timur Pocheptsov <Timur.Pocheptsov@digia.com>
AVFoundation on iOS lacks the ability to use AVCaptureVideoDataOutput and
AVCaptureMovieFileOutput simultaneously. Right now viewfinder stops working
as soon as we add movie file output. The only workaround
we have now is to write video/audio 'maually' - creating asset writer
and feeding it with audio/video samples.
Change-Id: I33a63546783279c545f0433b5051287269825d3f
Task-number: QTBUG-37655
Reviewed-by: Yoann Lopes <yoann.lopes@theqtcompany.com>
It was hardcoded to ARGB32, which is not a good idea, at least on iOS
where the necessary conversion is slow.
We now pick the QAbstractVideoSurface's preferred format, or if no
surface is set, we pick the default from AVFoundation.
As a result, the QML VideoOutput will now always use the NV12 format.
Change-Id: I65205c706455502883b8098f0b5c0577b4106e01
Reviewed-by: Timur Pocheptsov <Timur.Pocheptsov@digia.com>
When stopping, we don't actually stop the GStreamer pipeline, we just
pause it and prevent preroll frames from being shown.
We also need to make sure the last presented frame is cleared in that
case, otherwise it stays on screen.
Fixed for both 0.10 and 1.0.
Change-Id: Ibe26a7567f271ae0c3d8819eb9d35d6a95da1c6a
Reviewed-by: Christian Stromme <christian.stromme@theqtcompany.com>
- When the video capture resolution or the image capture resolution
changes, we now always set the viewfinder resolution to the highest
available one with the same aspect ratio as the capture resolution.
We were previously not doing anything if the new capture resolution
had the same aspect ratio as the current viewfinder resolution.
- Some devices don't support using a viewfinder resolution different
from the video capture resolution. Make sure we handle this case.
Change-Id: I8d3ab7b01c56ed78d1ca838a522ba459692fc332
Reviewed-by: Christian Stromme <christian.stromme@theqtcompany.com>
0df8d839 didn't fix all of the problems.
Task-number: QTBUG-46159
Change-Id: I4f2a390da1cca13775302aeef60c522e0da55a1a
Reviewed-by: Timur Pocheptsov <Timur.Pocheptsov@digia.com>
When missing, make check is an empty target, meaning the test
is never run in the CI.
Change-Id: I6f09e0f78c26b631fdfe89d2520dc0012d177843
Reviewed-by: Yoann Lopes <yoann.lopes@theqtcompany.com>
Padding rows & columns are included in the video frame that is sent to
the VIV direct texture. Scale the UV coordinates to ensure the padding
pixels aren't shown.
Change-Id: I8b870a95ff786f9f80f42f0cc33f468b0c9c3863
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
Reviewed-by: Yoann Lopes <yoann.lopes@theqtcompany.com>
We need a native X window handle to be able to use xvimagesink. If Qt
doesn't use the Xcb platform plugin, don't try to use that element.
This patch makes QVideoWidget usable when the system doesn't run X11
but xvimagesink is installed anyway. Widget and window controls become
invalid and the plugin therefore falls back to the renderer control.
Change-Id: I6efd410508f35b06f00f6e4e5149c97cbd280d0d
Task-number: QTBUG-46169
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
Maximum zoom value was incorrectly reset to 1.0 when doing the transition
Active -> Loaded -> Active.
Change-Id: I799900b1597637039d6c28d1d694fb6340b10540
Reviewed-by: Timur Pocheptsov <Timur.Pocheptsov@digia.com>
AVCaptureConnection.videoMaxFrameDuration is supported only since 10.9.
Task-number: QTBUG-46159
Change-Id: I8ea57b69e97ea3802b5c444c57ab090c4edf31e2
Reviewed-by: Timur Pocheptsov <Timur.Pocheptsov@digia.com>
The time was in milliseconds but should be in microseconds.
Task-number: QTBUG-45571
Change-Id: I54f07975e7a6233254a338dcde8075f740b5455c
Reviewed-by: Christian Stromme <christian.stromme@theqtcompany.com>
Also, crop the screenshot image to not show any window decoration,
and fix the example \brief.
Change-Id: I86310bdc5a4f124180eac1a1d27064286ba34f59
Reviewed-by: Venugopal Shivashankar <venugopal.shivashankar@digia.com>
Added missing documentation for Camera.FlashVideoLight and corrected
it for Camera.FlashTorch.
Change-Id: Ib1823b6554bd2f9db92e047fd4755ae7fb3e0f96
Reviewed-by: Venugopal Shivashankar <venugopal.shivashankar@digia.com>
fe13f5bb introduced a bug which could leave the active texture unit set
to GL_TEXTURE1 or GL_TEXTURE2. When the material is done updating its
state, it needs to make sure the active texture unit is reset to 0
since other materials might assume that's the current value.
Change-Id: I64a6a0cd9fcecdf0fa483909193f70045ff0e9fe
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
