Plane alpha was being sent as a float but read as unsigned, which, when
converted to a float was causing things that were supposed to be
translucent to be opaque.
Bug: 33739111
Bug: 33737738
Test: Scrim behind modal dialogs is translucent and there is no more
glitch when rotating apps
Change-Id: I6896bdbb9928bd1aa7265b3ef25a7a96d54ee227
Similar to IAllocator, introduce IComposerClient to manage resources.
Rework the interface such that most state changing calls are batched in a
"command buffer" and execute together. The goal is to reduce the number
of IPC calls needed to set up the composer.
Test: builds and boots
Change-Id: I324009243234c4d2482ca0ef2591377b11530fc9
Introduce IAllocatorClient to manage resources owned by a client (e.g., SF
or VTS). This makes sure there is no resource leak when SF or VTS
crashes.
This also fixes two unrelated bugs
- sizeof(Buffer) != sizeof(void*) on 32-bit impl.
- layerCount was not set to 1 in tests
Test: builds and boots
Change-Id: I67f5cdd64b97fb3ce1b931099c25f59cc8517f21
Remove duplicated types from allocator@2.0 and composer@2.1. Those types
are now in common@1.0.
Bug: 32238126
Test: builds and boots
Change-Id: Ibdb907e47969f5a5a98a14a2e2b77a9c5880beff
To allow genrules with more than one tool, rename the tool property to
tools and make it an array, replace $tool with $(location <label>),
and use $() for other variables for consistency.
Bug: 31948427
Test: compare build.ninja
Change-Id: I3d714f70a2af0dc60faeee10e09b6ed166601f1d
There is no need to use a genrule to copy headers, just move them into
an appropriate directory in the source tree and export them directly.
Bug: 31948427
Test: mmma -j frameworks/native/libs/ui
Change-Id: I9c5f385e29c78d47479c77a546cfc737e4dd40f8
Copied from system/graphics.h with two changes
- YCbCr -> YCBCR
- HAL_TRANSFORM_RESERVED is removed
Test: make
Change-Id: I1d343ff176ac61f911fee949861ce4b12255cd47
Since HALs might run binder services, we need to start thread pool for
both binder and hwbinder.
Bug: 32021609
Test: builds and boots
Change-Id: I9779e86b1e611a180b1984af36c417dafc3329bc
The default implementatoin is built on top of conventional gralloc1.
Test: make android.hardware.graphics.mapper.hallib
Change-Id: I8647dc69dbb1637a141d36358d5769f450422ac1
IMapper is loaded in-process and is defined in C++ instead of HIDL. To
guarantee binary compability, the interface is a struct of function
pointers instead of a class of pure virtual member functions. The
functions pointed to by the function pointers must also have C-linkage and
have all of their parameters PODs.
Implementations are expected to be installed to
/oem/lib/hw/android.hardware.graphics.mapper.hallib.so
/vendor/lib/hw/android.hardware.graphics.mapper.hallib.so
/system/lib/hw/android.hardware.graphics.mapper.hallib.so
(lib64 instead of lib for 64-bit implementations)
The loader will look for the symbol "HALLIB_FETCH_Interface" and use it to
fetch the interface.
Test: make android.hardware.graphics.mapper@2.0
Change-Id: I3a2c7a6ce18a90adda6df2d89559c957a122fdf0
The default implementatoin is built on top of conventional gralloc1.
This also adds a static library, libgralloc1-adapter. It is intended to
ease the porting of gralloc0 to gralloc1.
Test: booted to launcher, tested with YouTube and some games.
Change-Id: Id640b1d5a1e1eea1aafabb6c134e6be6e71afff5
It differs from gralloc1.h in that
- buffer descriptors are created from a struct, BufferDescriptorInfo, to
reduce round trips
- testAllocate is a function of its own
- buffer allocation and export are two different steps
- reference counting and buffer mapping are moved to gralloc-mapper
- BAD_HANDLE is renamed to BAD_BUFFER
- GRALLOC1_CONSUMER_USAGE_FOREIGN_BUFFERS is removed
- CPU_{READ,WRITE}_OFTEN no longer implies CPU_{READ,WRITE}
Test: make android.hardware.graphics.allocator@2.0
Change-Id: Ibe9367d5b1701c0e1009da829f27fed0f7d98828
