1. 29 Sep, 2017 - 1 commit
  2. 28 Sep, 2017 - 8 commits
  3. 26 Sep, 2017 - 6 commits
  4. 25 Sep, 2017 - 1 commit
    • Oliver Wolff's avatar
      winrt: Use QMutex instead of criticalSections · 020e7b4c
      Oliver Wolff authored
      
      critical sections are really unpredictable when it
      comes to order of activation. It was possible that
      the loop in QWinRTAbstractVideoRendererControl's
      syncAndRender blocked its setBlitMode for up to 30
      seconds on application start. During this time the
      camera screen just stayed black. The whole approach
      seems to work a lot better when QMutexes are used.
      
      Task-number: QTBUG-63015
      Change-Id: Ib1b0fa1da35fe299896068146254e4cf1d9616fb
      Reviewed-by: default avatarMaurice Kalinowski <maurice.kalinowski@qt.io>
      020e7b4c
  5. 22 Sep, 2017 - 1 commit
    • Yoann Lopes's avatar
      GStreamer: fix video output stopping when the main thread is blocked · f6838120
      Yoann Lopes authored
      
      When a new frame is ready to be rendered, our gst sink waits until the
      frame is actually processed in the GUI thread, but never more than 300
      ms. The time limit is there to avoid potential dead locks in specific
      situations. Before, if the wait would timeout, the sink would signal
      that there was an error rendering the frame, which would in turn put
      the pipeline in an error state and would stop processing any further
      frame. We now simply skip the frame if the GUI thread is blocked for
      too long and signal that everything went fine to the pipeline.
      This was already the logic in place for GStreamer 0.10 (see
      qvideosurfacegstsink.cpp).
      
      Task-number: QTBUG-60509
      Change-Id: I5173a15340c0e2065bb2fb5ca3bc045ac84ba7e1
      Reviewed-by: default avatarChristian Stromme <christian.stromme@qt.io>
      f6838120
  6. 21 Sep, 2017 - 1 commit
  7. 12 Sep, 2017 - 2 commits
  8. 08 Sep, 2017 - 1 commit
  9. 06 Sep, 2017 - 1 commit
  10. 31 Aug, 2017 - 3 commits
  11. 22 Aug, 2017 - 1 commit
    • Sami Nurmenniemi's avatar
      Enable tests for boot2qt · c341b2b2
      Sami Nurmenniemi authored
      The tests for boot2qt were disabled with commit
      b9bfe241
      
      . The tests can
      be enabled since qtdeclarative now fallbacks to software
      renderer if OpenGL is not supported.
      
      Two changes were needed to pass tests on boot2qt
      1. Add option to disable tests requiring WAV format and use
      that in the boot2qt environment
      2. Skip some tst_QPainterVideoSurface tests if GLContext is
      not supported by the platform
      
      Task-number: QTBUG-60268
      Change-Id: I236207fedb505124a2c7edf0caccd85011ab1416
      Reviewed-by: default avatarSimon Hausmann <simon.hausmann@qt.io>
      c341b2b2
  12. 17 Aug, 2017 - 1 commit
  13. 16 Aug, 2017 - 1 commit
  14. 08 Aug, 2017 - 1 commit
  15. 18 Jul, 2017 - 1 commit
  16. 12 Jul, 2017 - 2 commits
  17. 07 Jul, 2017 - 2 commits
  18. 06 Jul, 2017 - 3 commits
    • Edward Welbourne's avatar
      Use qDegreesToRadians() instead of hand-coding it · 0b0c0e72
      Edward Welbourne authored
      
      It have the virtue of documenting its meaning directly.
      
      Change-Id: I42f1be6b4c8e69dcf8b8711b42b19759a6aebe78
      Reviewed-by: default avatarAllan Sandfeld Jensen <allan.jensen@qt.io>
      0b0c0e72
    • Bartosz Golaszewski's avatar
      PulseAudio: fix a pthread_mutex crash in error path · 53538ff9
      Bartosz Golaszewski authored
      
      QSoundEffect (pulseaudio) uses a workaround for stability issues:
      although the pulseaudio mainloop mutex is recursive, it utilizes a
      separate lock counting. This is not the best solution, but it is a
      part of a larger set of changes which improved the stability in CI
      under heavy load.
      
      QSoundEffect always calls pa_threaded_mainloop_lock/unlock() from the
      same thread so the additional lock counting works in normal situation
      even though it doesn't use atomic types.
      
      However if pa_context_connect() fails, pa_threaded_mainloop_unlock()
      is called without regard to current lock count. This leads to random
      double-unlock aborts if pa_context_connect() fails more than once
      (e.g. after the reconnect scheduled from onContextFailed()).
      
      Fix this by always using the PulseDaemon wrappers around
      pa_threaded_mainloop_lock/unlock().
      
      Task-number: QTBUG-61725
      Change-Id: I41eb9a76892a6646fd5620ef8f686473b339464f
      Reviewed-by: default avatarChristian Stromme <christian.stromme@qt.io>
      53538ff9
    • Liang Qi's avatar
      Merge remote-tracking branch 'origin/5.9' into dev · 2282cd01
      Liang Qi authored
      Conflicts:
      	.qmake.conf
      
      Change-Id: I5f9bed6cb611471db61b9c69004f16786822625d
      2282cd01
  19. 30 Jun, 2017 - 3 commits