Age | Commit message (Collapse) | Author |
|
|
|
bug - 2563005
Change-Id: I79a2169022ffcddee3261bc2929f0bea28e50c5a
|
|
Change-Id: I38ce72815c61cdc1039b0ec27d48a004b0ba669c
|
|
We can run into problem where deallocate() called from two
separate thread which breaks PV's thread model where
every task is scheduled to run within a single omx thread.
I am pretty sure that this is why the free chunk available
notification is lost sometimes; however, this is certainly
not the fix to all the problems that we have seen so far.
bug - 2484098
Change-Id: I88bc494a11ffd739bc5771fa75b4643de503714f
|
|
the number of outstanding output buffer count. As a result, the media server either hangs if omx encoder node thinks it is still waiting for some buffers to be returned from the omx component, or silently crashes if the omx encoder node frees the buffer prematurely"
|
|
number of outstanding output buffer count.
As a result, the media server either hangs if omx encoder node thinks it is still waiting for some buffers
to be returned from the omx component, or silently crashes if the omx encoder node frees the buffer prematurely
bug - 2484098
Change-Id: Ic928a29ad1365c31b93f8a51942b4cf85445bd5d
|
|
server to crash
Currently, we hold the lock for the fragment queue for writing. This obviously resolves
the memory corruption problem, but could lead to performance issue.
TODO:
MOT is reviewing PV's implementation of vector to see whether we use a second lock to prevent
frames to be released while we using the first element of the queued frames to do file writing.
bug - 2501987
Change-Id: If25ffd93702f93c5ee4120beca234156c9405895
|
|
bug - 2484098
Change-Id: Ica88bc4905adecc2ee506fd2f5b9e2c0ecf01fe5
|
|
public constants, this mismatch prevented proper metadata extraction.
Change-Id: I1c1d541be8aebc7d45ec39891c53ba43fa946ece
related-to-bug: 2074137
|
|
|
|
Change-Id: Id6eb47e74a1c3bfaf935a3c8c4da7c7b0a34b625
|
|
are returned by the encoder in a single output buffer.
This patch also reverts most of the unnecessary changes introduced by QCOM
bug - 2489213
|
|
OpenCore's software decoders are all using 2 output buffers.
Holding more than 1 buffer in the video MIO can lead to starvation
and playback freeze.
bug - 2478420
|
|
When the number of bytes returned from audio driver is <= 0,
the existing audio MIO terminates the audio recording thread
in the MIO immediately. but when the application tries to
terminate the audio recording via stop or reset, the stop
or reset can potentially wait forever.
This is not verified, thus I enable the logging first to get
confirmation should this happens.
bug - 2484098
|
|
|
|
|
|
bug - 2484098
|
|
sholes
Add API to retrieve number of frames dropped by audio input kernel driver.
Submitted on behalf of Masaki Sato <masaki.sato@motorola.com>
|
|
The problem was that the pmemInfo can be changed from both the OpenCore omx proxy thread and the callback thread, and thus needs to be protected.
bug - 2374541
|
|
|
|
|
|
for doing Ogg Vorbis decompression)
|
|
|
|
Merge commit '031db8796b5140efb30b21671f32d12137989834'
* commit '031db8796b5140efb30b21671f32d12137989834':
Fix intermittent Mac build errors.
|
|
|
|
|
|
bug - 2317216
|
|
Previously the fragment writer thread queue was a static queue and if
the number of output buffers was greater than the queue size an overflow
would occur and encoded frames would be dropped, resulting in corrupted
video or missed audio.
- We will do performance optimization later.
- Improve the lock usage based on reviewer's comment
|
|
|
|
library for doing Ogg Vorbis decompression)"
This reverts commit 8a10af55f9b10ea77dc78fb221ebeda2204fb34e.
|
|
library for doing Ogg Vorbis decompression)"
|
|
Merge commit '5b11918b7a4571dffdb94d76d9d03104e9830813'
* commit '5b11918b7a4571dffdb94d76d9d03104e9830813':
android-2.1_r1 snapshot
eclair snapshot
|
|
|
|
for doing Ogg Vorbis decompression)
|
|
See http://b/issue?id=2391369
|
|
OpenCore searches in an URL (after the protocal portion) for ":", and "/".
Then, it uses the string after ":" to indicate the actual port number.
However, when ":" comes after "/", ":" does not indicate the port number.
bug - 2317216
|
|
|
|
related-to-bug: 2393577
|
|
|
|
|
|
platform-specific.
Merge commit '0bb93756e59a04933d4ab896ce593f994bf5267d'
* commit '0bb93756e59a04933d4ab896ce593f994bf5267d':
The number of frames that need to be hold in the video MIO is platform-specific.
|
|
We add an extra member variable to the video mio base class, and ask the
derived class to overwrite this value if necessary.
|
|
|
|
The primary directive is in
codecs_v2/audio/aac/dec/util/getactualaacconfig/Android.mk
Fixes the following warning:
build/core/copy_headers.mk:15: warning: overriding commands for target `out/target/product/sholes/obj/include/libpv/getactualaacconfig.h'
build/core/copy_headers.mk:15: warning: ignoring old commands for target `out/target/product/sholes/obj/include/libpv/getactualaacconfig.h'
Change-Id: I97c2f04fbe315deb725f15efebbbac040cd2e986
|
|
although a sw-based codec was found
|
|
is, when sw-based codec decoders are requested.
Thus, the request for sw-based codec becomes "preferred" request now, instead of "forced" ones.
|
|
frames instead of 1 so that SF can have more room to complete composition. bug 2309411
Merge commit 'afa07b52460a1c9f88294f14f5a2a259543c88b5' into eclair-mr2
* commit 'afa07b52460a1c9f88294f14f5a2a259543c88b5':
Fix tearing for 720P video playback.
|
|
Hold 2 decoded video frames instead of 1 so that SF can have more room to complete composition.
bug 2309411
|
|
MediaScanner specific to the PV implementation.
The encoding autodetection functionality now lives in libmedia as part of the MediaScannerClient implementation, as it is not PV specific.
|
|
(software based vs hardware based)
Merge commit '674995079ccb3544671ae00d378d2d932ca9d062' into eclair-mr2
* commit '674995079ccb3544671ae00d378d2d932ca9d062':
Enable the support for selecting preferred video codecs (software based vs hardware based)
|