…ed vs hardware based) This patch allows us to exclusively use software video codec for metadata retriever to avoid cases where the metadata retriever contents hw accelerated video codecs with some foreground video playback process. Currently, if the hw accelerated video codec is taken by the metadata retriever, then the foreground video playback process has to use software-based video codec for playback. End users won't have a very good experience using software-based video codec for playing high resolution videos in the foreground, if not always impossible. The change touches many files in OpenCore, but the actual change in each file is just a line or two. Thanks for comments from the reviewers. I was able to greatly reduce the number of files that are changed. We should have used the configuration files to advertise whether an omx core is hardware-based or software-based. Currently, this patch recognizes the software codecs by checking the component name. If the component name contains "PV", then it is a software-based codec. This is a hack. In addition, it would be nice to have more options for selecting a codec. For instance: option a. hardware only; b. hardware preferred; c. software only and d. software preferred. but that adds more complexity. Removed a new method based on reviewer's comment.
The problem was that when DoStop or DoReset was called, the first sample time stamp,iAudioFirstFrameTs, was reset right way, which made the VF detection calculation in the writeComplete() callback incorrect. The solution is to move the iAudioFirstFrameTs reset logic to DoStart() to make sure that we still always has the correct initialized first sample time stamp.
Add method to AndroidAudioInput to change audio source. Force AndroidAudioInput audio source to AUDIO_SOURCE_CAMCORDER if a video input is present.
…n in camera mio
…le HD (1280x720). Fix failures to generate thumbnails for HD contents.
There is a crash due to OMX_UseBuffer change. In Reset(), media input node deletes the media input node port, which deletes the allocator. When encoder node releases the media data, it calls the allocator to release the memory. That causes crash. Cherry-picked from MR2 branch. commit af748e6 Author: Jianhong Jiang <email@example.com> Date: Fri Oct 23 14:41:57 2009 -0700 RIO-7762: OMX_UseBuffer support buffers allocated in MIO. The change depends on https://android-git.corp.google.com/g/30984
… a call to OMX_SetParameter.
Also increased the size of iPrivateDataFsiFragmentAlloc to prevent from running out of space. This is a follow up for: commit 903c9e2 Author: Jianhong Jiang <firstname.lastname@example.org> Date: Fri Oct 23 14:41:57 2009 -0700 RIO-7762: OMX_UseBuffer support buffers allocated in MIO. The change depends on https://android-git.corp.google.com/g/30984
The change depends on https://android-git.corp.google.com/g/30984
by copying a string without its 0 terminator.
bug 2171732 This is just to help TI to track down the OMX component issues. Ideally, we still need to have a fix for this issue, since we must take out battery if this occurs.
…he clock is reset. http://b/issue?id=2171037
…c frame instead of IDR. http://b/issue?id=2167163 J.D. & D.S.
This reverts commit a371da5. http://b/issue?id=2164330 J.D. & D.S.