Hi Zhou-hao, On 30 March 2017 at 10:53, Zhuo-hao Lee <zhuo-hao....@intel.com> wrote: > Add an error handling to droid_swap_buffers() to handle the undefined > case of dri2_surf->buffer==null. > We found that when running CTS testAbandonRepeatingRequestSurface on > Chromebook, the test would fail. After digging, we found that the > Android framework relied on an undefined behavior to abort the camera > transmission. Therefore, we added an error handling to return > EGL_BAD_SURFACE to Android framework to force the camera transmission. > > the testAbandonRepeatingRequestSurface test flow: > (https://android.googlesource.com/platform/cts/+/android-7.1.1_r23/ > tests/camera/src/android/hardware/camera2/cts/RobustnessTest.java#921) > > testAbandonRepeatingRequestSurface { > 1. Create surface for the camera > 2. Setup camera parameter, callback function and underlying surface > 3. Setup repeating request to camera and start capture 60 frames > 4. Poll 60 frames from Camera and put to surface > 5. Release the surface texture <-- From now, Android framework will > call droid_swap_buffers() with dri2_surf->buffer=null > 6. Expect the camera's repeating request stop and get the last > frame number <-- Time out here because Camera isn't aborted > 7. Checking for the frame number > } > > The problem is happened after step 5 releasing the underlying surface > texture. The step 6 expects that the camere is aborted and should > trigger onCaptureSequenceCompleted() callback. This solution adds an > error handling which will return EGL_BAD_SURFACE to android framework > to force camera tramission be aborted. > One of the Google/CrOS developers proposed an alternative solution [1], which seem better imho. Perhaps you can help by reviewing or testing it ?
-Emil [1] https://patchwork.freedesktop.org/patch/147322/ _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
