* Qualcomm caused hell-on-earth with libavenhancements- an
apparently well-intentioned effort to minimize the amount of
change in f/av which has backfired. In order to use this
library, an OEM must not change *anything* *anywhere* in f/av
to maintain ABI exactness. And that ABI exactness is different
per SoC branch and constantly in motion. Best of all, the
base classes are monkey patched with a proprietary library which also
is different per SoC branch and much exactly match the f/av
code. And if you get the ABI right, then your customizations
stop working because they've been hijacked by overloading. The whole
message bus is hijacked, and you're wasting your time.
* Now that FFMPEG can talk to all the hardware codecs, there is no
reason to tolerate this. Our stack has more features, more codec
support, and it's open source all the way.
* I love Qualcomm's SoCs, but come on guys.. I expect every OEM
will complain about this AV shit then proceed to gut it as I am doing.
Change-Id: I8ed0de62d64cc10054933cd70a9958ea10359c7a
The variable persist.sys.usb.config is no longer honored by the framework,
and it's presence causes unnecessary toggling of the USB driver, which
disonnects ADB and makes the device's connection unstable.
Delete it.
Bug: 21404762
Bug: 18905620
Change-Id: Ibcc3b6a6f5523695d9daf01e53e8f5ed71ad19fe
* We have updated component and we should just kang bacon's media config
to be closer to caf versions omx compponents..
* Basicially the mobvation was that wmv3/vc1 support was missing.
Change-Id: Ibc29445b2ea935459b0c6e9faf76b6e6a648a71b
We now get the qcom-common SELinux policy. Get rid of our old policy
and add back config as we need it, rather than trying to merge the two.
Change-Id: I5ca5098c653f09c7901343d0ae71793f6eb21ab8
-Also clean up code and get rid of warnings.
Clean up from Arne Coucheron works on other samsung wrappers.
-Requires device tree to include a CameraParametersExtra.h
for the specific device.
Change-Id: Icab1b9ec17710861f735a05dcdd87dac932c63c2
The IR pulse sequences are written to a local buffer to be
passed to the kernel.
Previously, we doubled the fixed size buffer (from 1K to 2K)
to work around a crash that was occurring for long sequences.
Another crash has been found when using Smart IR Remote and the
POWER function of the TX P42X10E remote by Panasonic (TV category).
Rather than double the size again and wait for another counter-
example, dynamically allocate the buffer and grow it as needed
to handle any IR sequence.
Change-Id: Id4fb799d8a5ad4242e04a68e6cc6436a14f4d2d0
* Requires a proprietary component and does not work well on 8974
anyway. Remove it to pass MediaCodecList CTS test.
Change-Id: I0b98f07fadef3e63842597109ea7da1a37ccea8e
According to:
https://developer.android.com/reference/android/hardware/ConsumerIrManager.html#transmit%28int,%20int[]%29
the pattern is supposed to be in ms. Apparently Samsung and HTC
in versions 4.4.2 and earlier of Android were treating this as
the number of pulses instead of ms. This is what our HAL is
supporting because it mimic'ed what the vendors were doing.
However, we should switch to the correct spec of the pattern (ms)
and eventually this will cause apps to converge on the right
patterns.
Currently, this change will break half the existing apps and cause
the other half of the apps to stop being broken and to start
working.
Change-Id: I87cedcc8e0b85d7b96ed63c7d4f82cc53ecb4ecb
