- We only support 0, 90, 180, and 270 degree clockwise rotation
- Some players are known to ignore composition matrix in the MP4 file,
although this is part of the MP4 file standard.
Both QT and YT are supporting the rotation
Change-Id: I1b7f66a801e9d9c49d889c9b06dd6173fa7e76c4
Fix premature release of recording frames when physical address or metadata is stored in input video buffers
- bug 3158459
Change-Id: If297189d2a87fc3abfda68c29ac75b490b30a902
- add a sniffer for DRM files
- add DRMSource and DRMExtractor for es_based DRM
- add pread in FileSource.cpp for container_based DRM
- add native DRM framework API calls in the player for
DRM audio/video playback
Change-Id: I4b9ef19165c9b4f44ff40eeededb9a665e78a90f
o Do not count the reserved space for moov if the meta data size is small
o Do not count the extra 1KB disturbing small file estimation.
o Reduce the default minimum reserved space from 4 KB to 3 KB.
o Estimate the moov size based on both duration AND file size limit is set
and set it to the smaller estimated value.
low risk change
bug - 3111983
Change-Id: I6ac2adb979d8cc12d6b4f1813d000c989add0199
Added a method to expose the audio session id at AudioSink interface
so that the AudioPlayer in stagefright can retrieve it.
Also:
- Fixed audio effect send level not being initialized in mediaplayer.
- Fixed compilation error when LOGV is enabled in mediaplayer JNI
Change-Id: I4bb55454fd63d646e0e677692d737c4843fb05fb
The problem was that even though user does not explicitly request the max file size
limit via MediaRecorder.setMaxFileSize(), the file writer sets an implicit file
size limit if 32-bit file offset is used on user's behalf. The reserved free space
is estimated based on the file size, if the file size limit is set by the user.
The fix is to add an extra bool to tell the difference between an
explit requested file size and an implicit file limit and use that
to set the estimated moov box size accordingly.
Change-Id: I731aca6c7833aa764ed7b905edb77721577471b3
commit 29a4d3effb05a2e074cb0693316ab1977baeb0b6
Author: Andreas Huber <andih@google.com>
Date: Mon Sep 27 12:01:32 2010 -0700
Fully working implementation of MPEG2TSWriter (for AAC and AVC sources).
Change-Id: I8a32a47565b647bf6c078c520e39565e08ea0d84
commit f4dec4c3899f3be393508e180d6c07e249d3335e
Author: Andreas Huber <andih@google.com>
Date: Mon Sep 27 10:36:31 2010 -0700
More reliable identification of MPEG2 transport streams. Don't keep scanning forever in case the stream does not have both audio and video tracks.
Change-Id: Icc5b4e8be145b2805e8776559546a6818342aea7
commit 4fe3cc942f9b3d3cf54138b828c41214aa916dd2
Author: Andreas Huber <andih@google.com>
Date: Mon Sep 27 08:23:39 2010 -0700
test code
Change-Id: I16560a17661407d06497f99ff88230724bb898af
commit 64d988b24f49f179a90fa677be11c823959e734b
Author: Andreas Huber <andih@google.com>
Date: Thu Sep 23 14:42:52 2010 -0700
First shot at supporting writing to an MPEG2 transport stream.
Change-Id: Ie537939a99fa3ddc0c7661c47c18277584817c74
Change-Id: If78fd034af8f6e8ceac8dbeff96d5ecb3f6b96dc
Changed type of decay time, reverb delay and reflections delay parameters
from signed to unsigned int to match OpenSL ES interface definition.
Also fixed some type casts in lvm reverb wrapper.
Change-Id: I5ca5e76a87c2590f01f031f3168355586ef22556
When the recorded file becomes large, the metadata size can
no longer be ignored. This makes it possible to save the
recorded file when the storage becomes almost full at the
end of the recording session.
Change-Id: Ief038080f825c9946ce550949c03e914aec1e31a
The problem was that the time to receive an output buffer
from an audio encoder is different because the encoder does not
need to read from the source for all output buffers. This leads
to large fluctuation in terms of wall clock duration between two
neighboring audio sample outputs from the audio encoder. As a
result, the media time for the video track after adjustment using
the drifting changes wildly sometimes.
This patch addresses this issue by only updating the media drift
time when an audio source input buffer is read. the wall clock
for the audio track is also calculated at the same time when
the input audio buffer is read at AudioSource.
bug - 2959800
Change-Id: I3174aa182f744784b540f0a7198524d4eee8bd7b
Modified lvm reverb wrapper code to expose a preset reverb interface.
Also removed debug log from bundle and reverb wrapper.
Change-Id: If9b95d91e25a6ff834decdfdda34b17df9b46967
o Make the API consistent with SF framework, which the MediaSource
provides a return status for stop
o Also, helps to convey errors that occurred right when a
premature stop() is called, leading to a potentially
mal-formed output file.
Change-Id: I52a932345f38570fdf8ea04d67d73dd94ccd30ef
This implementation uses fixed points instead of floating points. It
is slightly inaccurate compared to the old one but still perfect for
visualization purpose. It runs 40% faster on passion, 5 times faster
on sholes, and of course 14 times faster on sapphire.
Change-Id: I1e868417bcffda091becf106a7b941d02813faec
o Only do this for realtime applications
o Adjust other track clock based on audio clock
o Assume other track uses wall clock as the media clock
o Use some heuristics to reduce the size of stts box by 2/3.
- also
o Remove one unused key from MetaData.h
Change-Id: Ib9432842627b61795b533508158c25258a527332
