| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
| |
Pick-to: dev 5.15 5.12
Change-Id: Ic670af0c1e8ef7962bddda0f0a3b7f547ae25e69
Reviewed-by: Jani Heikkinen <jani.heikkinen@qt.io>
|
|
|
|
| |
Change-Id: Ibd51c84967f81fb4aef70173300bbf96c842e9c5
|
|
|
|
|
|
|
|
|
|
|
| |
GValue can contain allocated memory. Not unsetting it can cause memory
leak. This patch adds g_value_unset() calls to various places that miss
ones.
Change-Id: I78e0f8f6c558ada0e7828c11094132359c579c2a
Reviewed-by: Val Doroshchuk <valentyn.doroshchuk@qt.io>
(cherry picked from commit 24ac478e50dab189b53c749b34971a807aa6da4d)
Reviewed-by: Qt Cherry-pick Bot <cherrypick_bot@qt-project.org>
|
|\
| |
| |
| | |
Change-Id: I486735873cc2be7e21b2e58a0d014726b8b43873
|
|/
|
|
|
|
|
|
|
|
|
| |
+ 67cab3e01d4d26f95af2e202ad3fd16722b1c6c7 tst_QMediaObject: Remove checking of expected elapsed time
+ c22f263c5d7bd602197215978d59e53d67a7dd6d Skip all tst_QMediaPlayerBackend tests when there is no codec available
+ 61a6a09429cf8c34a356748a3bb76fc7868a54b1 ALSA: Calculate buffer size based on min period time
+ 0f2e2dba7e728596db2c7207032f04a9e1a82333 Bump version
+ 76f5027181f690d6d0311f3d9f10061da7166ac9 Vivante: Keep only one frame
Change-Id: I5aa94f8d4cebdb3192d993d904a483064f713c69
Reviewed-by: Paul Wicking <paul.wicking@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The imx6 vivante implementation of videonode used by qtvideosink
holds reference to two frames during rendering.
It releases the first frame only when it receives the third frame.
This makes jerky playback when frames are not coming at constant rate.
It displays the previous frame during video playback.
Change-Id: If306bdcd358fac46a1fd0000ccb068424aee813a
Reviewed-by: Karim Pinter <karim.pinter@qt.io>
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
(cherry picked from commit f48e16dc45cde8f634823c40aa573b18673cf13d)
|
|
|
|
| |
Change-Id: Idf57d46aed6b5a0a5a9f9a36adf5498ee710554d
|
|\
| |
| |
| | |
Change-Id: I58208af8152fccfc292ea6b4c4796a46fdb25502
|
| |
| |
| |
| |
| |
| |
| |
| | |
+ 1e50366ea7cd8909f049f161a41afbd152a94579 AVF: Always lock the capture device when start the camera
+ 62d70affd1f0d70dad2507feaa9caa3ab0cef877 Bump version
Change-Id: I426aea6e0b19a6f8ad204d846a5e76ba39f4a539
Reviewed-by: Paul Wicking <paul.wicking@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
When defualt buffer size (100000us) and period time (20000us)
are out of range, currently buffer size is computed with
respect of either 2 or 4 chunks of maximum buffer time.
Which is not enough.
Fixing computing of buffer time based on minimum period time:
How many minimum periods could fit in maximum suggested buffer time.
Fixes: QTBUG-43256
Task-number: QTBUG-79526
Change-Id: I722e5145e3b6d323e306da420ec764db3575225f
Reviewed-by: Christian Strømme <christian.stromme@qt.io>
(cherry picked from commit 8b92454d6fffcee7a8cf1fb5e0c92920c77733c8)
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Previously tst_QMediaPlayerBackend::isWavSupported() always returned true
unless boot2q env, which is not correct.
Since if there is no codec installed in CI (but it is not boot2qt),
also need to skip the tests.
Change-Id: Ib58864bb3648a3d8f6c99354034f4ad85d5e4f98
Fixes: QTBUG-82673
Reviewed-by: Asmo Saarela <asmo.saarela@qt.io>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
(cherry picked from commit 060e2e7846441a65bfdab5cf0afea008b1bbb13f)
Reviewed-by: Paul Wicking <paul.wicking@qt.io>
|
|/
|
|
|
|
|
|
|
|
| |
It is often failed on CI when no enough resources.
Task-number: QTBUG-82582
Change-Id: Ibcddab73e6dfd24c3ab064ed8ea89843ebdc92ea
Reviewed-by: Paul Wicking <paul.wicking@qt.io>
(cherry picked from commit d5d42a24b5b9d2079d584defac70aef1a1feadef)
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
|
|\
| |
| |
| | |
Change-Id: If08a53f5b981cfc47a4755603f7088e14e6eab37
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
+ 49ce2d783b1bb74c119794420f1e96e2fe0531c0 Bump version
+ d381fb9823dac7e1e19b2b2af6b57ddf1306346a Qt Multimedia - fix build for iOS
+ 0144393242078421217f1967e64f00e3e32b898a AVF: Don't apply settings if the camera is not active yet
Change-Id: I433c8723272e5e3f53ae4fbdbacd5707e0153ac3
Reviewed-by: Paul Wicking <paul.wicking@qt.io>
|
| | |
|
|/
|
|
|
|
|
|
|
|
|
|
| |
If the viewfinder settings are valid, need to lock the capture device
to avoid reseting settings to default values.
See also a8123e737140719549252806e0e6a1c121359f79
Fixes: QTBUG-81048
Change-Id: I834815ef6c5ef28d8017d04bdb9d5256c02d1de7
Reviewed-by: Paul Wicking <paul.wicking@qt.io>
(cherry picked from commit 265bfb22e7f0f5333e584f6dd91926b8516e394e)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In macOS, a capture session can still automatically configure the capture
format after you make changes.
If frame rates are applied to the capture device and the camera
is still not active, these rates will be overridden by old values
when the capture session will be started.
For this purpose lockForConfiguration is currently used within startRunning method
of capture session. But in case if the settings are already applied to the capture device,
we don't call lockForConfiguration within starting of the capture session
(since it is needed to be called only when the settings have been changed).
Suggesting to postpone setting of the format (to the capture device) until
the camera is started.
This will lead to apply settings before startRunning method with proper configuration lock.
Fixes: QTBUG-81048
Change-Id: I04664e7b63474ce28571e888e524170b995a38d6
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
(cherry picked from commit a8123e737140719549252806e0e6a1c121359f79)
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
|
|\
| |
| |
| | |
Change-Id: Ifa15362ba556e8af191db386fb0dd843f1ddf868
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
+ 012affe3194a48d8bb10d0a100f1d1bbb24ecf7e DirectShow: Fix crash when there is no surface on flush()
+ 94852a47c130c3dc265d900390bf0d7b53cbab3e DirectShow: Add startTime and endTime to QVideoFrame
+ 9e96fe2cbf79a44039de0edcc854050d84b87588 DirectShow: Round stop position down to available bytes in IAsyncReader
+ e04b67b45672d529adaa785876ecbbbcef445d0f Bump version
Change-Id: I16b55a36c22492d49388b52f53e3342a090d05c1
Reviewed-by: Christian Strømme <christian.stromme@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Since RestrictedBool conversion operator was introduced, objective-C
method calls require explicit .data() on our smart pointers.
Fixes: QTBUG-78685
Change-Id: Ie32ebf0442cd56a01466aa58b89651d7c293de6d
Reviewed-by: VaL Doroshchuk <valentyn.doroshchuk@qt.io>
(cherry picked from commit b64bdf32ce10b039e8d7b1be18d001e89f315615)
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
|
|/ |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Currently requested IMediaSample might contain the end time that exceeds the
available bytes which causes returning an error in WaitForNext() and
stopping the playback.
Regarding to IAsyncReader::Request documentation:
The start and stop positions should match the alignment that was decided when the pins connected.
The stop position might exceed the real duration.
If so, the method rounds the stop position down to the actual alignment.
Fixes: QTBUG-77782
Change-Id: I644e25bfc6bb8f6d345b8424b79fb56490d82c0e
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
|
|\ |
|
|/|
| |
| |
| | |
Change-Id: I6137c73fe14e171aaa184e25ebc3a8374e4d91e2
|
| |\
| | |
| | |
| | | |
Change-Id: If27d8e3bb898607c5bf07dbfc97476859ae24646
|
| | |
| | |
| | |
| | |
| | | |
Change-Id: I282469f6c634283cec565b692b29b16cf0fdbea4
Reviewed-by: VaL Doroshchuk <valentyn.doroshchuk@qt.io>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
If the application seeks to a new position, the Media Session restarts
the presentation clock at the specified seek time.
The media source might deliver samples with a slightly earlier time stamp
if the seek time does not fall on a key-frame boundary.
And IMFSample->GetSampleTime will return time that is related to
a position when the clock has been restarted, and can be also negative.
So if it needs to have times from the beginning, it would require to add seeked position
to times from IMFSample.
It also reverts c4de056a6aa44567cdbf2ce91a464e597ad4af8f
Fixes: QTBUG-77849
Change-Id: I1a7cb7bd18aee73087a61d2ed2c3d644ad0fbd50
Reviewed-by: Christian Strømme <christian.stromme@qt.io>
|
| |/
|/|
| |
| |
| |
| |
| |
| |
| |
| | |
When QDeclarativeRendererBackend is being destroyed, it clears the
surface and postpones releasing IMediaControl, which is done on worker
thread. It also calls flush() where null surface is used.
Fixes: QTBUG-77829
Change-Id: I327583c1f8fb7585dbec3c3fb7e80d0155cc4819
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
|
|/
|
|
|
| |
Change-Id: I1ed22d646c2ff230217e1ecefaf37a0a45fa4b54
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
|
|
|
|
|
|
|
|
|
| |
The feature was added in 5.12.2.
And it is relevant only for GStreamer backend.
Change-Id: I87e22e506158dc61bb7f111e74806e7eb8c28d7a
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Reviewed-by: Paul Wicking <paul.wicking@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
ResolutionPredicate defines an order on the resolutions (expressed as
QSize) of AVCaptureDeviceFormats, from smallest (width as primary,
height as secondary sort key) to largest, as the lexicographical
less-than of width and height. A a lexicographical order over Strict
Weak Orders is a Strict Weak Order of the product type. So far, so
good.
The logical negation of a Stict Weak Order is, however, not a Strict
Weak Order (not-less-than is greater-or-equal, not greater-than), so
it cannot be used as the predicate in std::sort.
Rewrite the ResolutionPredicate as an adapter that can be used with
std::less _or_ std::greater (or even std::equal_to), piggy-backing on
std::tuple to implement the lexicographical sort for us, then replace
not2(ResolutionPredicate) with ResolutionPredicate<std::greater>.
Rename the predicate to something more apt.
This also solves the use of deprecated (and in C++20, removed)
std::not2.
Change-Id: I6f81b149e53a5b4299b188bf3ce996f638bf3334
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
|
|
|
|
|
|
| |
Change-Id: Ie1c77fa373d37979e69c08ab4a959c18efc9599e
Fixes: QTBUG-76135
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We promised to not provide any video frames before pause() or play() is called.
Regardless the media is loaded or not.
It should work like following:
setMedia(new)
// ...
// no video frames returned yet
pause()
// prerolled frame should be shown
To implement this, i.e. showing the prerolled frame after pause, we do seeking to the beginning.
But it is totally not necessarily when play() is requested,
since here is new media and playback will start from the beginning anyway.
This also produces a hang/stuck on some embedded systems with custom pipelines or playing from qrc.
So decided to show prerolled frame and this means to seek to the beginning only when pause() is called.
Task-number: QTBUG-65399
Change-Id: I9b3dc632fa0df4a1115d852c2d480fd5c7fd9ee5
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
|
|
|
|
|
|
|
| |
Change-Id: I8b6b0f9b278ebb4b3eff0caaedab62cebfe9a695
Fixes: QTBUG-77097
Reviewed-by: Rolf Eike Beer <eb@emlix.com>
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
| |
qtvideosink element is available in pipelines only if QVideoRendererControl
is used, because it requires a video surface.
So if qtvideosink is used inside pipeline but a surface is not available,
created null/dummy surface to render video frames to and show
a warning about this.
Change-Id: I924d3baca994363550b7920176e29e9cd0c4dd1f
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
|
|
|
|
|
|
|
|
| |
CARenderer was not re-created so black frames were rendered in video if current OpenGL context changed.
Task-number: QTBUG-77270
Change-Id: I8df423aaa9239c938363141cca1ae4136cde839b
Reviewed-by: VaL Doroshchuk <valentyn.doroshchuk@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Currently QAudioRecorder is implemented to find a codec and muxer that match some caps.
E.g. if we would want "audio/x-wav" container with PCM format, there is only one gst element
which supports it:
wavenc element accepts "audio/x-raw", "audio/x-alaw" or "audio/x-mulaw"
and provides buffers in "audio/x-wav" or "audio/x-rf64".
Next, we would need to find a codec which provides buffers in
"audio/x-raw", "audio/x-alaw" or "audio/x-mulaw". It might be mulawenc and alawenc.
And finally we combine them together to pipeline:
audio/x-raw -> mulawenc|alawenc -> wavenc -> audio/x-wav
And as a result there will be wav file and encoded in "audio/x-alaw" or "audio/x-mulaw"
and not in desired PCM format.
So suggesting to add audioconvert element to the audio encoders to allow to link
with various containers:
audio/x-raw -> audioconvert -> wavenc -> audio/x-wav
Change-Id: Icd77124267f5a0999625fa1fe4e3de3740218741
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
|
|
|
|
|
|
|
| |
I might happen when shareContext remains the same but current context has been changed few times.
Change-Id: I55a7a9eab03776e3d1ca6d32117de6714692bbf0
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Since starting the surface could be done after updating the geometry of
VideoOutput element (paint event is received),
and this leads setting invalid geometry to renderer and never updates by correct one.
Because the geometry is not "dirty" anymore, means already set.
Need just to update geometry when (or keep trying before) the surface is already started with
proper video surface format.
Change-Id: I338d7fe355c20f2027c6231241714376e9b569b2
Fixes: QTBUG-76205
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In case of clearing current playlist by passing nullptr to QMediaPlayer::setPlaylist, empty url is passed to a backend:
QMediaContent m(playlist, QUrl(), false);
setMedia(m);
The status should be updated by QMediaPlayer::NoMedia if empty url is
passed.
Fixes tst_QMediaPlayerBackend::playlistObject()
Task-number: QTBUG-65574
Change-Id: Iea03e3fcb8d74c9b1b482aa06b42686cf8c9e9a7
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Moved stream types to QGstCodecsInfo and added possibility to fetch audio
or video codecs by a container. Or fetch containers by audio or video codec.
It would allow to debug supported codecs and containers:
QGstCodecsInfo containers(QGstCodecsInfo::Muxer);
QGstCodecsInfo audioCodecs(QGstCodecsInfo::AudioEncoder);
for (auto &container: containers.supportedCodecs())
qDebug() << audioCodecs.supportedCodecs(containers.supportedStreamTypes(container));
Change-Id: I26bf5579db6974a166d408c4865a9ffe314e3e15
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
|
|
|
|
|
| |
Change-Id: I9f811b4d0aadb04d3552dcd1f26a1565dfc230e2
Reviewed-by: VaL Doroshchuk <valentyn.doroshchuk@qt.io>
|
|
|
|
|
| |
Change-Id: I5c2dadfe1154961adcecf779126166aea4d32393
Reviewed-by: VaL Doroshchuk <valentyn.doroshchuk@qt.io>
|
|
|
|
|
| |
Change-Id: Ied89175c4b7f5df090deac174b1da1e8496ee533
Reviewed-by: VaL Doroshchuk <valentyn.doroshchuk@qt.io>
|
|
|
|
|
| |
Change-Id: I9d74355fb800234d4fbd12235b89965921027223
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
|
|
|
|
|
| |
Change-Id: Ia50ce0d2c94b3b064e9cd571e598ecd37ee2b5e3
Reviewed-by: VaL Doroshchuk <valentyn.doroshchuk@qt.io>
|
|
|
|
|
|
|
|
|
|
| |
GetFrameRateList passes an unmanaged pointer to the caller which must be
manually freed with CoTaskMemFree. Additionally the Chromium project
notes that some drivers cause quirky return values which we would not
catch without stricter checks. See: https://chromium.googlesource.com/chromium/src/media/+/8cc93abd7339eeb9b7c2a12cca07b3dc245b2139/video/capture/win/video_capture_device_win.cc#484
Change-Id: I6aa4a6ea1ac0241e585e98cf9ff63240bacd3956
Reviewed-by: Christian Strømme <christian.stromme@qt.io>
|
|
|
|
|
|
| |
Fixes: QTBUG-76816
Change-Id: I2909c2f2df91ac26b5104e24892310aa62cad172
Reviewed-by: Christian Strømme <christian.stromme@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
QSample objects live and handled in loading thread (which uses QNetworkAccessManager).
When the app is finished and going to be destroyed, all static objects are destroying as well.
In case if static QNetworkConfigurationManagerPrivate (which is used by QNetworkAccessManager)
is destroyed before static QSampleCache, and loading of the resource is not finished yet
(still executing QNetworkAccessManager::get()), this produces a crash.
Since the loading thread is started only when loading of new QSample is requested,
(and all events are also handled by this thread)
proposing a fix to wait before loading thread is finished when a sample is requested to be released.
This postpones deleting of the QSample either when new sample is requested to load
or when QSampleCache is destroyed.
This makes sure that no loading thread exists when all QSoundEffects objects
and afterwards QNetworkConfigurationManagerPrivate are already deleted.
Change-Id: I55669ea4c2796a48cae4f0465f7f74d89e393675
Fixes: QTBUG-76090
Reviewed-by: Christian Strømme <christian.stromme@qt.io>
|