By making IComponent present itself as an IInputSink, the wrapper code
can be made to execute in the same process as the IComponent instance,
hence reducing an IPC hop in the case where an InputSurface comes from a
different process.
Test: make cts -j123 && cts-tradefed run cts-dev -m \
CtsMediaTestCases --compatibility:module-arg \
CtsMediaTestCases:include-annotation:\
android.platform.test.annotations.RequiresDevice
Bug: 126296132
Change-Id: I61409ccfa9a9d6b152e2c9363ff119302ce8ae1a
This includes the following changes:
1. Revert commit 7570e3cd4b.
2. Revert commit 2a40d309b7eefe36d2f78bbff42aafad24eb65b0.
3. Use the new bufferpool (@2.0).
4. Use the old bufferqueue (@1.0).
5. Add IInputSink to eliminate downcasting from IBase.
Test: Builds
Bug: 112362730
Change-Id: I2b5f2e2821c62cbb7166320e1e128d883e278480
This CL makes transitioning of Codec2 HAL from private to public
smoother. The CL will be reverted later.
Test: Builds
Bug: 112362730
Change-Id: Id0e199112e968e35ed3b27413baae810f1ecb2df
The purpose of this CL is to make transitioning smoother. This CL
will be reverted later.
Test: Builds
Bug: 112362730
Change-Id: I65bf589facd0dda7a9af8b552eb795153dc1bd87
In order to synchronize message id when a new connection is made,
message Id is added as a return value. Also the last posted message Id
is given as paramter, when observer is notfied of new messages.
Bug: 112203066
Change-Id: Ic3921fd176aec1abc7de3ea88c730ad46182d28f
Bufferpool clients are notified bufferpool events via FMQ messages.
Usually the messages are processed by pipeline via piggybacking. To
ensure process FMQ messages even if the process is in idle state,
Observer hidl interface is added. Buffer invalidation caused by
pipleline completion uses Observer interface in order to invalidate
buffers after all normal operations are finished.
Bug: 112203066
Change-Id: I2a4a172d78c6bcedcc809c559321929ebc91c30d
Buffer Invalidation FMQ is designed as a shared FMQ between clients in
order to minimize memory consumption. In that case, regarding FMQ is
unsync FMQ in order to not to block writer in any situation.
Bug: 112203066
Change-Id: I58989ccd332ba790193007f43dd66ecbd617f484
BufferPool HAL 2.0. This is identical to 1.0 HAL now.
The update will be applied afterwards.
Bug: 112203066
Change-Id: I8f5bbf979116f66357f2bc0f8114697472adf762
The tests insisted on presence of codecs as well as functionality, but
the list it reads is a superset of what may be on the device. Modified so
that if a codec is not present, we ignore it rather than failing the test.
The Vts tests that check OMX Store are being augmented to ensure any
advertised codecs are indeed present.
Bug: 80029024
Bug: 80530946
Test: vts run VtsHalMediaOmxV1_0Host
Change-Id: Icd6d049c5c0d31f47ef3dbda52d6927e38bd894d
BufferPool HAL for allocating, deallocating, and recycling buffers
on behalf of media.stagefright.
Bug: 72651719
Change-Id: I30bf630e41de2c8718aaaf8cbf5ee3f48f2859bc
timestampDevTest flag can be used to disable timestamp deviation
tests. The flag is not working as intended. This is corrected.
Test: make vts -j99 BUILD_GOOGLE_VTS=true TARGET_PRODUCT=aosp_arm64 \
&& vts-tradefed run commandAndExit vts \
--skip-all-system-status-check --primary-abi-only \
--skip-preconditions --module VtsHalMediaOmxV1_0Test -l INFO
Bug: 64743270
Bug: 65166112
Merged-In: I611e96498a4677b6ccbff33c291d56eed7b48f6f
Change-Id: I611e96498a4677b6ccbff33c291d56eed7b48f6f
(cherry picked from commit dc6270ecc1)
Components can send various kinds of port settings changed events all
at once. During a full port reconfiguration this is causing problems
due to limited flexibility in the message handling design. For now,
before committing to a full port reconfiguration defer any events
waiting in the queue to be addressed to a later point.
Test: make vts -j99 BUILD_GOOGLE_VTS=true TARGET_PRODUCT=aosp_arm64 \
&& vts-tradefed run commandAndExit vts \
--skip-all-system-status-check --primary-abi-only \
--skip-preconditions --module VtsHalMediaOmxV1_0Test -l INFO
Bug: 64468705
Merged-In: I537cdd3e7d92cdc6d54168f091d7897afe541599
Change-Id: I537cdd3e7d92cdc6d54168f091d7897afe541599
(cherry picked from commit a230bad8d4)
The eColorFormat field in OMX_PARAM_PORTDEFINITIONTYPE structure
has no effect on the component. This has to be configured from
OMX_VIDEO_PARAM_PORTFORMATTYPE. This is corrected. This color format
information is passed to graphic buffer during its allocation
Bug: 63796949
Merged-In: I51b1e59631647d1291ecfac373781ac5e5a0ffd5
Change-Id: I51b1e59631647d1291ecfac373781ac5e5a0ffd5
(cherry picked from commit 1441864f4c)
