OSDN Git Service
Android (Google) Code Review [Thu, 9 Jul 2009 00:21:05 +0000 (17:21 -0700)]
Merge change 6558 into donut
* changes:
Add timestamps to camera frames
Dave Sparks [Thu, 9 Jul 2009 00:17:26 +0000 (17:17 -0700)]
Add timestamps to camera frames
James Dong [Wed, 8 Jul 2009 22:06:51 +0000 (15:06 -0700)]
Treating PVMFInfoContentTruncated as an error event since the playback session cannot be recovered once the event occurs
bug
1968334
Jianhong Jiang [Wed, 8 Jul 2009 15:56:53 +0000 (08:56 -0700)]
RIO-6947: Retrieve PLATFORM from system properties for user agent string.
Sahil Sachdeva [Tue, 7 Jul 2009 21:38:16 +0000 (14:38 -0700)]
Follow-up - Renaming OMX Master Core methods, to fix build break because of missing
changes in Player, Author and FMU Test apps.
Sahil Sachdeva [Mon, 6 Jul 2009 22:19:32 +0000 (15:19 -0700)]
1850776 Fix Valgrind issue in Mp4ParserNode - Mismatched free() / delete / delete []
Sahil Sachdeva [Thu, 25 Jun 2009 21:07:44 +0000 (14:07 -0700)]
Rename OMX master core methods to avoid conflicts (link problems)
Jianhong Jiang [Tue, 30 Jun 2009 21:35:31 +0000 (14:35 -0700)]
Remove _TEST_AE_ERROR_HANDLING from media input node.
Jianhong Jiang [Sat, 27 Jun 2009 00:13:52 +0000 (17:13 -0700)]
RIO-6871: Fix the Donut RTSP UserAgent string
Android (Google) Code Review [Fri, 26 Jun 2009 16:29:28 +0000 (09:29 -0700)]
Merge change 5482 into donut
* changes:
Fix crash in PMEM heap destructor due to sequencing issue. Bug
1919195.
Dave Sparks [Fri, 26 Jun 2009 02:27:59 +0000 (19:27 -0700)]
Fix crash in PMEM heap destructor due to sequencing issue.
Bug
1919195.
Android (Google) Code Review [Fri, 26 Jun 2009 00:06:31 +0000 (17:06 -0700)]
Merge change 5387 into donut
* changes:
When OpenCore 2.0 was integrated, it broke the audio stream MIO. This patch signals the observer so that playback starts correctly. Bug
1837526
Android (Google) Code Review [Thu, 25 Jun 2009 21:56:40 +0000 (14:56 -0700)]
Merge change 4836 into donut
* changes:
OMX M4V encoding drops the first I frame
Sahil Sachdeva [Fri, 19 Jun 2009 22:52:04 +0000 (15:52 -0700)]
OMX M4V encoding drops the first I frame
Dave Sparks [Thu, 25 Jun 2009 19:11:19 +0000 (12:11 -0700)]
When OpenCore 2.0 was integrated, it broke the audio stream MIO.
This patch signals the observer so that playback starts correctly.
Bug
1837526
Jianhong Jiang [Wed, 24 Jun 2009 20:30:08 +0000 (13:30 -0700)]
NJ-1425: Add AMR set bitrate support.
Dave Sparks [Wed, 24 Jun 2009 01:11:29 +0000 (18:11 -0700)]
Modify camera MIO to use new camera callback interface.
This removes the discrete callbacks and replaces them with two common
callback methods.
Bug
1884362
James Dong [Fri, 19 Jun 2009 18:33:25 +0000 (11:33 -0700)]
Don't access queue element without checking its size (bug
1922537)
Jianhong Jiang [Fri, 19 Jun 2009 21:12:51 +0000 (14:12 -0700)]
AStyle change for OpenCORE.
Sahil Sachdeva [Fri, 19 Jun 2009 00:24:52 +0000 (17:24 -0700)]
1850579 Fix the valgrind errors reported when running Simulator
Sahil Sachdeva [Thu, 18 Jun 2009 00:18:58 +0000 (17:18 -0700)]
Follow-up of NJ-1422 : Fixing PlayerDriver and AuthorDriver to catch any leaves
which happens in player or author, during playback or recording. This is to avoid
a leave to get transferred to upper layers, resulting in calling abort and crashing.
Android (Google) Code Review [Thu, 18 Jun 2009 00:28:41 +0000 (17:28 -0700)]
Merge change 4526 into donut
* changes:
Log null author command in authordriver.cpp (bug
1922537) Note this does not resolve the root cause for the mediaserver crash reported in bug
1922537: 1. why was a null author command received from the author engine (PV is looking into this issue)? 2. why was the time stamp out of the order (QCOM and PV are looking into this issue)? 3. why did we receive multiple event callbacks for a single state transition request (QCOM is looking into this issue)?
James Dong [Wed, 17 Jun 2009 23:24:08 +0000 (16:24 -0700)]
Log null author command in authordriver.cpp (bug
1922537)
Note this does not resolve the root cause for the mediaserver crash reported in bug
1922537:
1. why was a null author command received from the author engine (PV is looking into this issue)?
2. why was the time stamp out of the order (QCOM and PV are looking into this issue)?
3. why did we receive multiple event callbacks for a single state transition request (QCOM is looking into this issue)?
Sahil Sachdeva [Wed, 17 Jun 2009 00:46:44 +0000 (17:46 -0700)]
[
1841509] Fails to play the audio with the attached video
For the content, size of the audio config buffer exceeds the size of the audio config data
because of which the extra bits were taken as a part of next raw aac stream, which resulted in
bit stream corruption.
Fix, force to consume all bits of the audio config buffer.
James Dong [Mon, 15 Jun 2009 20:09:21 +0000 (13:09 -0700)]
Don't ignore compiler warnings (bug fix:
1315153)
Jianhong Jiang [Mon, 15 Jun 2009 15:45:43 +0000 (08:45 -0700)]
http://b/issue?id=
1267355
NJ-1090 : disable HEAD request for progressive streaming or download.
Marco Nelissen [Thu, 11 Jun 2009 22:22:56 +0000 (15:22 -0700)]
Perform some checks on the Xing header, and reject it if it looks funny,
which causes the code fall back to another code path to determine the
length of the file. This makes it so that NPR podcasts don't show up
as being 13 seconds long.
Marco Nelissen [Thu, 11 Jun 2009 22:19:48 +0000 (15:19 -0700)]
Revert "Perform some checks on the Xing header, and reject it if it looks funny,"
This reverts commit
2c6f649b82e91bda20648083b3b4cef25b7c792c.
Marco Nelissen [Thu, 11 Jun 2009 18:19:58 +0000 (11:19 -0700)]
Perform some checks on the Xing header, and reject it if it looks funny,
which causes the code fall back to another code path to determine the
length of the file. This makes it so that NPR podcasts don't show up
as being 13 seconds long.
Android (Google) Code Review [Thu, 11 Jun 2009 01:00:30 +0000 (18:00 -0700)]
Merge change 3821 into donut
* changes:
Send Reset() command to PV player engine no matter CancelAllCommands() succeeded or not.
James Dong [Wed, 10 Jun 2009 23:43:50 +0000 (16:43 -0700)]
Send Reset() command to PV player engine no matter CancelAllCommands() succeeded or not.
Android (Google) Code Review [Wed, 10 Jun 2009 20:47:20 +0000 (13:47 -0700)]
Merge change 3680 into donut
* changes:
Don't export DSO symbols from camera mio and audio mio for authoring. The code size reduction is about 12 k for libopencore_author.so.
Android (Google) Code Review [Wed, 10 Jun 2009 19:20:01 +0000 (12:20 -0700)]
Merge change 3734 into donut
* changes:
Cleanup value add in makefile.
Nicolas Catania [Wed, 10 Jun 2009 18:17:18 +0000 (11:17 -0700)]
Fix the stop handling when the writer thread is used.
The flush operation is not the only way to stop the composer.
The other way is by reaching a limit (size or time), in which case
an event is reported to the author engine AND the file is flushed.
This is an issue because the flush call on the writer thread now needs
to be reentrant (not called from the main thread anymore).
To move the call back on the main thread we use an extra field to hold
the event that should be fired. On the next fragment, the event is sent
and the DoStop method will be run from the main thread as a result.
Added extra code to make sure the worker thread flushes its buffer queue.
Added a bool to remember that exit was called. exitPending form Thread
gets cleared before the dtor is invoked.
In the flush method, don't wait if the flush happened.
James Dong [Wed, 10 Jun 2009 01:45:16 +0000 (18:45 -0700)]
Don't export DSO symbols from camera mio and audio mio for authoring.
The code size reduction is about 12 k for libopencore_author.so.
Nicolas Catania [Wed, 10 Jun 2009 17:58:38 +0000 (10:58 -0700)]
Removed trailing whitespaces. Untabify.
Jianhong Jiang [Wed, 10 Jun 2009 17:44:50 +0000 (10:44 -0700)]
Cleanup value add in makefile.
Marco Nelissen [Tue, 9 Jun 2009 16:32:13 +0000 (09:32 -0700)]
Fix more opencore simulator breakage. Hopefully that's the last of it.
Marco Nelissen [Tue, 9 Jun 2009 15:57:37 +0000 (08:57 -0700)]
Fix opencore simulator breakage
Jianhong Jiang [Tue, 9 Jun 2009 02:53:08 +0000 (19:53 -0700)]
Export more symbols to make GCC 4.4 happy.
By Jing Yu.
Android (Google) Code Review [Tue, 9 Jun 2009 00:48:27 +0000 (17:48 -0700)]
Merge change 3508 into donut
* changes:
Don't bother adding unsupported metadata to our list, and reject album art over 3 megabytes. This should go a long way towards the media scanner not crashing on corrupted mp3s.
Marco Nelissen [Tue, 9 Jun 2009 00:17:32 +0000 (17:17 -0700)]
Don't bother adding unsupported metadata to our list, and reject album art over 3 megabytes.
This should go a long way towards the media scanner not crashing on corrupted mp3s.
Android (Google) Code Review [Mon, 8 Jun 2009 22:58:24 +0000 (15:58 -0700)]
Merge change 1547 into donut
* changes:
NJ-1409: (platform/external/opencore) Add support for set sampling rate and number of channels
Jianhong Jiang [Mon, 8 Jun 2009 21:44:18 +0000 (14:44 -0700)]
NJ-1409: (platform/external/opencore) Add support for set sampling rate and number of channels
Android (Google) Code Review [Mon, 8 Jun 2009 21:16:59 +0000 (14:16 -0700)]
Merge change 3383 into donut
* changes:
OpenCORE code size reduction.
Jianhong Jiang [Mon, 8 Jun 2009 19:14:03 +0000 (12:14 -0700)]
OpenCORE code size reduction.
James Dong [Fri, 5 Jun 2009 01:25:33 +0000 (18:25 -0700)]
Bug
1897865:
The final solution is not ready for OTA.
This change temporarily marks off the changes for spinning off a separate file writer thread.
Android (Google) Code Review [Wed, 3 Jun 2009 21:47:41 +0000 (14:47 -0700)]
Merge change 2998 into donut
* changes:
RIO-6494: OMX node can go into fatal error without notifying the engine properly. This can result ANR.
Jianhong Jiang [Wed, 3 Jun 2009 02:37:00 +0000 (19:37 -0700)]
RIO-6494: OMX node can go into fatal error without notifying the engine properly. This can result ANR.
Android (Google) Code Review [Wed, 3 Jun 2009 02:04:45 +0000 (19:04 -0700)]
Merge change 2817 into donut
* changes:
NJ-1422 :- Native crash after record the 1 min video After 1min of recording AuthorEngine issues an internal Stop() which Stops the MediaInputNode port and queues Stop on MIO component. MIO component still has some data queued in its queue and keeps sending data to MediaInputNode. MediaInputNode port does a OsclLeave with OsclErrNotReady since it has moved to Stopped State. Audio MIO does not handle this leave and leave is transferred to upper leaves, resulting in Abort. Fixed Audio MIO to handle leave properly by adding TRY CATCH block around writeAsync call. TBD - We need to also make sure that leave is caught in PlayerDriver and AuthorDriver rather thank transferring it to upper layers, so that we can avoid call of abort and avoiding a crash.
Sahil Sachdeva [Mon, 1 Jun 2009 18:34:43 +0000 (11:34 -0700)]
NJ-1422 :- Native crash after record the 1 min video
After 1min of recording AuthorEngine issues an internal Stop()
which Stops the MediaInputNode port and queues Stop on MIO component.
MIO component still has some data queued in its queue and keeps sending
data to MediaInputNode. MediaInputNode port does a OsclLeave with OsclErrNotReady
since it has moved to Stopped State. Audio MIO does not handle this leave and leave is
transferred to upper leaves, resulting in Abort. Fixed Audio MIO to handle leave properly
by adding TRY CATCH block around writeAsync call.
TBD - We need to also make sure that leave is caught in PlayerDriver and AuthorDriver rather
thank transferring it to upper layers, so that we can avoid call of abort and avoiding a crash.
Android (Google) Code Review [Wed, 3 Jun 2009 00:41:04 +0000 (17:41 -0700)]
Merge change 2902 into donut
* changes:
Use a worker thread to write the fragment to the sdcard.
Android (Google) Code Review [Tue, 2 Jun 2009 23:53:47 +0000 (16:53 -0700)]
Merge change 2873 into donut
* changes:
Fix broken open() call.
Nicolas Catania [Tue, 2 Jun 2009 07:30:40 +0000 (00:30 -0700)]
Use a worker thread to write the fragment to the sdcard.
Added a worker thread running at normal priority to write the
fragments on the sdcard.
Fragments are stored in a circular buffer.
There is flush method on the worker to ensure all data has
been written.
Dave Sparks [Mon, 1 Jun 2009 23:26:32 +0000 (16:26 -0700)]
Fix broken open() call.
Dave Sparks [Fri, 29 May 2009 01:40:49 +0000 (18:40 -0700)]
Use setVideoSurface call on MIO when surface changes
Dave Sparks [Fri, 29 May 2009 01:19:03 +0000 (18:19 -0700)]
Allow setVideoSurface to be called multiple times. Allow a null
to be passed so that video can continue playing without a surface.
Dave Sparks [Thu, 28 May 2009 16:39:10 +0000 (09:39 -0700)]
Allow subsequent calls to setVideoSurface after prepare()
Sahil Sachdeva [Fri, 22 May 2009 22:16:04 +0000 (15:16 -0700)]
Player Driver and Author Driver crash fix, if InitializeThread fails
Dave Sparks [Fri, 22 May 2009 02:08:32 +0000 (19:08 -0700)]
Route audio source from AuthorDriver::setAudioSource() to audio MIO.
This is a continuation of the changes needed to route the audio
source to allow developers to choose an audio source other than
the microphone.
Bug
1846343
Marco Nelissen [Wed, 20 May 2009 16:02:18 +0000 (09:02 -0700)]
Re-apply part of perforce CL 131021, which was lost with a recent integrate from
Marco Nelissen [Wed, 20 May 2009 14:56:05 +0000 (07:56 -0700)]
Re-apply perforce CL 134361-p9, which was lost with a recent integrate from
Android (Google) Code Review [Tue, 19 May 2009 17:41:39 +0000 (10:41 -0700)]
Merge change 1964 into donut
* changes:
RIO-6460: any pause/resume in the last audio frame will cause AMIO stuck.
Jianhong Jiang [Tue, 19 May 2009 16:54:18 +0000 (09:54 -0700)]
RIO-6460: any pause/resume in the last audio frame will cause AMIO stuck.
Dave Sparks [Sat, 16 May 2009 00:27:57 +0000 (17:27 -0700)]
Enable AGC and noise suppression for MediaRecorder
Bug
1725543
Dave Sparks [Thu, 14 May 2009 15:51:30 +0000 (08:51 -0700)]
Improve audio quality of recordings by increasing AMR bitrate.
Android (Google) Code Review [Wed, 13 May 2009 16:59:39 +0000 (09:59 -0700)]
Merge change 1512 into donut
* changes:
fix a tiny class design flaw in pvmf_omx_videodec_node.h change the access modifier to private for those methods that do not need to be visible outside of the class
James Dong [Wed, 13 May 2009 02:07:28 +0000 (19:07 -0700)]
fix a tiny class design flaw in pvmf_omx_videodec_node.h
change the access modifier to private for those methods that do not need to be visible outside of the class
Android (Google) Code Review [Tue, 12 May 2009 21:34:50 +0000 (14:34 -0700)]
Merge change 1423 into donut
* changes:
Fix OMX enc node memory leak. See also: https://android-git.corp.google.com/g/1396
Jianhong Jiang [Tue, 12 May 2009 01:23:11 +0000 (18:23 -0700)]
Fix OMX enc node memory leak.
See also:
https://android-git.corp.google.com/g/1396
Android (Google) Code Review [Tue, 12 May 2009 18:27:28 +0000 (11:27 -0700)]
Merge change 1415 into donut
* changes:
NJ-1410 : Fix for large time taken to start playback after seek on youtube contents
Sahil Sachdeva [Mon, 11 May 2009 23:55:51 +0000 (16:55 -0700)]
NJ-1410 : Fix for large time taken to start playback after seek on youtube contents
Android (Google) Code Review [Tue, 12 May 2009 01:27:39 +0000 (18:27 -0700)]
Merge change 1421 into donut
* changes:
Don't ignore exceptions in MediaScanner. The MediaScanner skips a directory when an error occurs while scanning the directory. This is reasonable if a file can't be read or something or a directory is not accessible. However, if a Java exception occurs, calling into Java again with a pending exception will cause the VM to be aborted. This change aborts processing and passes the Java exception back up to Jave for processing instead of continuing.
Dave Sparks [Tue, 12 May 2009 01:06:36 +0000 (18:06 -0700)]
Don't ignore exceptions in MediaScanner.
The MediaScanner skips a directory when an error occurs while scanning
the directory. This is reasonable if a file can't be read or something
or a directory is not accessible. However, if a Java exception occurs,
calling into Java again with a pending exception will cause the VM to be
aborted. This change aborts processing and passes the Java exception
back up to Jave for processing instead of continuing.
Android (Google) Code Review [Mon, 11 May 2009 22:34:05 +0000 (15:34 -0700)]
Merge change 1295 into donut
* changes:
RIO-6419, NJ-1416 Mp3 Progress Bar Hang Fix
Android (Google) Code Review [Mon, 11 May 2009 17:15:32 +0000 (10:15 -0700)]
Merge change 1301 into donut
* changes:
oscl tls registry race condition bug fix
James Dong [Fri, 8 May 2009 23:56:37 +0000 (16:56 -0700)]
oscl tls registry race condition bug fix
If the tls registry is removed in a thread while another thread is accessing an item in the registry, this causes crashes
- the registry was not protected at all, and now protected with a lock
- there is no need for an individual lock on the key; and thus it is removed
- called unlock before return for getInstance() and registerInstance()
Android (Google) Code Review [Mon, 11 May 2009 16:58:00 +0000 (09:58 -0700)]
Merge change 1162 into donut
* changes:
OC2 oscl registry race condition The idea is to only initialize the singleton table for once and only once and keep it around forever. As a result, the lock for locking the whole singleton table is not needed and thus removed in this change. This approach works only if the client program follows the protocol for using the singleton: it must call initialize() first and then call the rest methods in this class.
Sahil Sachdeva [Fri, 8 May 2009 21:50:40 +0000 (14:50 -0700)]
RIO-6419, NJ-1416 Mp3 Progress Bar Hang Fix
James Dong [Thu, 7 May 2009 18:33:37 +0000 (11:33 -0700)]
OC2 oscl registry race condition
The idea is to only initialize the singleton table for once and only once and keep it around forever.
As a result, the lock for locking the whole singleton table is not needed and thus removed in this change.
This approach works only if the client program follows the protocol for using the singleton:
it must call initialize() first and then call the rest methods in this class.
- changed the pointer to a singletontable to be a static object so that it is initialized statically
- removed an extra lock on the whole singletontable, since we don't need it
- moved initialize and cleanup to the header file for optimization. actually, these two methods can be removed. I put the removal of
these two methods as a FIXME.
Jianhong Jiang [Fri, 8 May 2009 18:24:13 +0000 (11:24 -0700)]
Fix the 2way build.
Marco Nelissen [Mon, 4 May 2009 20:35:49 +0000 (13:35 -0700)]
We were logging media scanner failures, but not which file failed, which is not very useful.
James Dong [Fri, 1 May 2009 07:23:18 +0000 (00:23 -0700)]
Incorporate all the changes in external/opencore in git master branch
- add two patches for fixing thumbnail extraction and authoring compatibility
- remove 2way engine from system image
Not done yet (pending one fix from PV)
James Dong [Fri, 24 Apr 2009 22:51:59 +0000 (15:51 -0700)]
Bug fix(
1807910): a crash in media recorder due to the use of locked camera
Sorry for the repeatedly review request because I messed up the git repo.
- change the signature of SetCamera() method to return a status code so that
we can check the status of SetCamera() in authordriver.cpp
- change camera MIO (android_camera_input.cpp) to use create factory method
for constructing a Camera client from a camera remote so that the status
of the construction can be verified before we allow recording to proceed.
- change some LOGD statements to remove compiler warnings due to the use
of %p for ssize_t variable "offset".
- change the call flow in authordriver.cpp to avoid code duplication
- fixed a bug in the previous change of the call flow
Jean-Baptiste Queru [Thu, 23 Apr 2009 00:56:53 +0000 (17:56 -0700)]
merge cupcake into donut
Dave Sparks [Wed, 1 Apr 2009 18:10:21 +0000 (11:10 -0700)]
AI 144029: Don't send write complete until frame is no longer displayed
We seem to be overwriting video frames while they are being displayed.
This change holds back an acknowledgement for the frame currently
displayed until the next one is received. The final frame is released
when EOS, stop, reset, flush, or cancel command is received.
BUG=
1749950
Automated import of CL 144029
Dave Sparks [Wed, 1 Apr 2009 18:09:25 +0000 (11:09 -0700)]
AI 144031: am: CL 144029 Don't send write complete until frame is no longer displayed
We seem to be overwriting video frames while they are being displayed.
This change holds back an acknowledgement for the frame currently
displayed until the next one is received. The final frame is released
when EOS, stop, reset, flush, or cancel command is received.
Original author: davidsparks
Merged from: //branches/cupcake/...
Automated import of CL 144031
James Dong [Wed, 25 Mar 2009 23:38:59 +0000 (16:38 -0700)]
Automated import from //branches/donutburger/...@142712,142712
James Dong [Wed, 25 Mar 2009 22:09:11 +0000 (15:09 -0700)]
Automated import from //branches/cupcake/...@142601,142601
Dave Sparks [Wed, 25 Mar 2009 06:05:41 +0000 (23:05 -0700)]
Automated import from //branches/donutburger/...@142513,142513
Dave Sparks [Wed, 25 Mar 2009 06:05:19 +0000 (23:05 -0700)]
Automated import from //branches/cupcake/...@142512,142512
Andreas Huber [Wed, 25 Mar 2009 04:24:45 +0000 (21:24 -0700)]
Automated import from //branches/donutburger/...@142071,142071
Niko Catania [Wed, 25 Mar 2009 03:57:28 +0000 (20:57 -0700)]
Automated import from //branches/donutburger/...@141819,141819
James Dong [Wed, 25 Mar 2009 02:02:50 +0000 (19:02 -0700)]
Automated import from //branches/donutburger/...@140956,140956
Andreas Huber [Wed, 25 Mar 2009 01:57:22 +0000 (18:57 -0700)]
Automated import from //branches/cupcake/...@142067,142067
Niko Catania [Wed, 25 Mar 2009 01:26:16 +0000 (18:26 -0700)]
Automated import from //branches/cupcake/...@141816,141816
Niko Catania [Wed, 25 Mar 2009 00:50:54 +0000 (17:50 -0700)]
Automated import from //branches/donutburger/...@140524,140524
The Android Open Source Project [Fri, 20 Mar 2009 06:08:53 +0000 (23:08 -0700)]
auto import from //branches/cupcake_rel/...@141571
The Android Open Source Project [Thu, 19 Mar 2009 00:39:44 +0000 (17:39 -0700)]
auto import from //branches/cupcake_rel/...@140373
The Android Open Source Project [Fri, 13 Mar 2009 20:04:21 +0000 (13:04 -0700)]
auto import from //branches/cupcake_rel/...@138607