hongzuo liu authored 2 years ago and Simão Gomes Viana committed 1 year ago

Incorrectly dropped event because of pending overdue app switch when the App switch key event has arrived.

https://issuetracker.google.com/u/1/issues/255902047



Change-Id: If6dd28a8793fe6f06b6100f2cc44eafa07c0f085
Signed-off-by: Simão Gomes Viana <devel@superboring.dev>

Adithya R authored 2 years ago and Simão Gomes Viana committed 1 year ago

Some vendors purposely disable brightness capability
even though the HWC completely supports it, in order
to leverage their custom brightness implementations.
One such example of this is in stock HWC of some
xiaomi devices.

Since light HAL offers lesser granual control than
HWC (since its in gamma space and not float), and
since it would be rather difficult to patch the blob
to return the right capabilities, allow forcing the
usage of HWC brightness functions via a prop:

ro.sf.force_hwc_brightness=1 (default 0).

Change-Id: I7d2fef03afbbb936572c717f769f30c9e49213e0
Signed-off-by: Adithya R <gh0strider.2k18.reborn@gmail.com>
Signed-off-by: Simão Gomes Viana <devel@superboring.dev>

Rajat Yadav authored 3 years ago and Simão Gomes Viana committed 1 year ago

Reset validateWasSkipped variable in the PostCommit phase of
draw cycle with Skip Validate.

CRs-Fixed: 3035058
Change-Id: I85348dd0c80bd1258ac630b000bd694b34af6e12
Signed-off-by: Simão Gomes Viana <devel@superboring.dev>

Rheygine Medel authored 2 years ago and Simão Gomes Viana committed 1 year ago

CRs-Fixed: 3215802
Change-Id: I92613a41a0cb9da2ad44e8a2ac803a5533ea6495
Signed-off-by: Simão Gomes Viana <devel@superboring.dev>

Adithya R authored 2 years ago and Simão Gomes Viana committed 1 year ago

Layer dataspace is initialized as unknown by default, unless
overriden by the buffer source. We don't require expensive
rendering for color conversion when the dataspace is unknown.

Change-Id: I079c520f63a65c77ba3162664656e607eafff991
Signed-off-by: Simão Gomes Viana <devel@superboring.dev>

Simão Gomes Viana authored 2 years ago and Simão Gomes Viana committed 1 year ago

Signed-off-by: Simão Gomes Viana <devel@superboring.dev>

dianlujitao authored 4 years ago and Simão Gomes Viana committed 1 year ago

Author: Bruno Martins <bgcngm@gmail.com>
Date:   Wed Oct 14 23:45:14 2020 +0100
Edit:   Adapt to new lineage soong config

Change-Id: Ic0f314f4053628667a921951f610839f36a5079c
Signed-off-by: Simão Gomes Viana <devel@superboring.dev>

Rheygine Medel authored 2 years ago and Simão Gomes Viana committed 1 year ago

Change-Id: Iecd2be3eb8775855d82763f069a38bb70fe53d4e
CRs-Fixed: 3229968
Signed-off-by: Simão Gomes Viana <devel@superboring.dev>

Rheygine Medel authored 2 years ago and Simão Gomes Viana committed 1 year ago

Currently, the properties are used as follows:
1. debug.sf.enable_hwc_vds - allows IDs to be generated for VDs
2. vendor.display.vds_allow_hwc - allows WFD to use HWC path

With this change, HWC path is enabled for WFD with only the vendor
property set. All other virtual displays require both of the
properties enabled.

Change-Id: Iab2c8d15d2c1cf24be0d371af8892c346634507f
CRs-Fixed: 3204941
Signed-off-by: Simão Gomes Viana <devel@superboring.dev>

Michael Bestas authored 3 years ago and Simão Gomes Viana committed 1 year ago

* These changes are part of CAF AOSP merge commits

Change-Id: I390f874347d259fca4429c19711be6c85b104090
Signed-off-by: Simão Gomes Viana <devel@superboring.dev>

Rajat Yadav authored 3 years ago and Simão Gomes Viana committed 1 year ago

Allow HWC based virtual display only for wifi display

CRs-Fixed: 2984439

Change-Id: Ifbb094a2a0101171b475bd5a60660a3599dce5ff
Signed-off-by: Simão Gomes Viana <devel@superboring.dev>

Chris Gross authored 4 years ago and Simão Gomes Viana committed 1 year ago

Bug: 169788930
Change-Id: Ic50f5675d0cd48f84fb9ff14221355741dc52129
Signed-off-by: Simão Gomes Viana <devel@superboring.dev>

Ramakant Singh authored 4 years ago and Simão Gomes Viana committed 1 year ago

1. SurfaceFlinger: Fix virtual display related issues. 1) Validate
output
buffer usage bit appropriately Validate output buffer usage bit only
against GRALLOC_USAGE_HW_COMPOSER to differentiate GLES only
conposition
and HWC/MIXED composition. 2) Exclude video encoder usage for
scratch
buffer Sink and Scratch buffers in VDS are using same usage
flags. This
causes video encoder flag to be set for scratch buffers
Also. Exclude
video encoder flag from scratch buffer usage flags as
scratch buffers
are used only as write back input and not video
encoder input.

2. sf: Allow VDS to use HWC -- Preserve VDS layer
pixel format based on
GRALLOC flags. -- skip color layer for vds

3. sf: Add secure content support for VDS

4. sf: Enable GPU comp. for non-primary displays
Ensure that "Return
status" of dequeueBuffer() complies with API
	requirements. Client relies
on this status to call requestBuffer()
	upon reallocation.

5. sf: Enable UBWC on virtual display scratch buffer Set
	GRALLOC_USAGE_HW_FB usage flag on virtual display Scratch
	buffer to
	enable UBWC.

Change-Id: Ia49a175372ca187a295531e18f8e84dc22a19486
CRs-Fixed: 2656027
Signed-off-by: Simão Gomes Viana <devel@superboring.dev>

Adithya R authored 2 years ago and Simão Gomes Viana committed 1 year ago

Taken from CLO (QSSI 13). Some Qualcomm devices can still
benefit from disabling backpressure propagation by setting:

debug.sf.disable_backpressure=1

[adi8900]: Adapted to Android U

Change-Id: I2346a1b314666706e1299b295f13b2cef6e00b4d
Signed-off-by: adi8900 <adrianszymanski242@gmail.com>
Signed-off-by: Simão Gomes Viana <devel@superboring.dev>

daniml3 authored 3 years ago and Simão Gomes Viana committed 1 year ago

The FOD layers should be always composed by the device to set the
custom zpos bits to the kernel. By default all layers are set to
device composition and the proprietary libsdmextension.so moves
some to client composition. Unfortunately that also affects the
Udfps touched layer. After this commit the layer below the Udfps
touched layer is forced to be client composition. For unknown
reasons libsdmextension will not change the Udfps touched layer
to client composition anymore.

tests:
- Run 'adb shell dumpsys SurfaceFlinger' and ensure that the Udfps
  touched layer is composed by the device
- Ensure that the Udfps touched layer correctly sets the zpos bits
  on the kernel while / after using WFD

Signed-off-by: daniml3 <danimoral1001@gmail.com>
Signed-off-by: Arian <arian.kulmer@web.de>
Change-Id: I8aeb98d18557ad4e971eaba74700ceb3058273ab

Pierre-Hugues Husson authored 4 years ago and Simão Gomes Viana committed 1 year ago

 * Supports changed z Udfps order
 * Supports changed Udfps usage bits

TheScarastic: Adapt to extension lib and support 12 BiometricPrompt
ArianK16a: Adapt layer name to UdfpsControllerOverlay for Android 13
    and only compare the prefix. Drop the BufferQueueLayer change
    because it seems unnecessary.

Co-authored-by: TheScarastic <warabhishek@gmail.com>
Change-Id: Id95aa73e06b4223a6b4f05c69fa2fc494f9a97b1

This change improves the performance of the WindowInfosListenerInvoker work done on SurfaceFlinger's background executor thread. The primary optimization made is not sending a WindowInfosReportedListener with every call to WindowInfosListener.onWindowInfosChanged. Instead, we send a new interface, WindowInfosPublisher, and a unique listener id to listeners when they're added.  Listeners call WindowInfosPublisher.ackWindowInfosReceived with their id after processing each update.

From traces taken during development, the new code is a major improvement, taking about 15% of the time spent previously on SurfaceFlinger's background thread for sending window infos. Performance with this change seems roughly in line with the performance in T.

Bug: 290377931
Test: atest WindowInfosListenerTest
Test: atest WindowInfosListenerInvokerTest
Test: manually killing system server and checking valid state on restart
(cherry picked from commit acd2258a)
(cherry picked from commit e8a7ab25)
(cherry picked from https://googleplex-android-review.googlesource.com/q/commit:82f3463d449eb13e28c5dbffeee16e10721c71d2)
Merged-In: Ib39ba935727df0bc1ab4030bcfe8301de7e64805
Change-Id: Ib39ba935727df0bc1ab4030bcfe8301de7e64805

This adds a unit test and fix for the bug reported by libfuzzer.
Changes made:
 * Expose GPUService as testable code.
 * Update main_gpuservice.cpp to use the new GpuService now located at
   gpuservice/GpuService.h
 * Make initializer threads members of GpuService
 * Join the threads in destructor to prevent heap-use-after-free.
 * Add unit test that waits 3 seconds after deallocation to ensure no
   wrong access is made.

Bug: 282919145
Test: Added unit test and ran on device with ASAN
(cherry picked from commit 3c00cbc0)
(cherry picked from https://googleplex-android-review.googlesource.com/q/commit:a8186037a730699b53e49de7241280c7532e5fc4)
Merged-In: I4d1d2d4658b575bf2c8f425f91f68f03114ad029
Change-Id: I4d1d2d4658b575bf2c8f425f91f68f03114ad029

LOG(FATAL) crashes the device, which is very severe for the user.

The accessibility hover event stream has consistency issues that we
are actively fixing, but these fixes have not yet had time to soak.
Instead of cherrypicking the fix, instead we can demote the severity
of the log so that errors do not crash the device.

Bug: 286037469
Test: atest InputDispatcherTest#InvalidA11yHoverStreamDoesNotCrash
Test: Use talkback, repeatedly enter touch exploration while also
      tapping with another finger. Eventually observe the failure
      message which is now an ERROR instead of crashing the device
      with a FATAL.
(cherry picked from https://googleplex-android-review.googlesource.com/q/commit:2f99cdb9150461c6b77c29b31cbf5792a9303c95)
Merged-In: Iceea086345691cdad8ca5e092629de094d666de1
Change-Id: Iceea086345691cdad8ca5e092629de094d666de1

Some HAL implementations can't support setLayerBuffer multiple times to
clear the per-layer buffer caches. Therefore, default this behavior to
disabled, and allow HALs to explcitily enable this behavior to obtain
the necessary memory savings.

Test: play videos with both true and false on both HIDL and AIDL
Bug: 285561686
(cherry picked from https://googleplex-android-review.googlesource.com/q/commit:f70d9c4528888134e16003470b76dd1e81b8631d)
Merged-In: I928cef25e35cfc5337db4ceb8581bf5926b4fbe3
Change-Id: I928cef25e35cfc5337db4ceb8581bf5926b4fbe3

Uncaptured touchpads (with DeviceMode::POINTER) still on the old stack
(i.e. the Sony DualShock 4 touchpad) were having touches ignored if they
started at coordinates that would be outside of the bounds of the
touchscreen. Since touchpads use relative motions and don't directly
relate raw touch locations to screen locations, this was incorrect, and
caused touches to be ignored with certain screen/touchpad dimension and
orientation combinations.

Bug: 280396539
Test: connect Sony DualShock 4 by USB, rotate screen to portrait, check
      touches starting in all areas of the pad are turned into pointer
      movements
Test: atest inputflinger_tests
Change-Id: I9a92ba79246ee8fc25418c2e248ae2c3839a0d58

Cap the max amount of HDR headroom beyond which the
layer is clipped. This avoids over-dimming the SDR content
range in the resulting SDR screenshot

Test: screenshot of silkfx
Bug: 286942637
Change-Id: I67df32d9c192da18df516972ed92e9bd64063186

SensorService::threadLoop() "continues" if there are no real sensor
events, which skips the handling of runtime sensor events.

Fix: 281452823
Test: m
Change-Id: I0fe9fe9a7baa19ce9c8cb32d36e736ac83d2fe52

Currently, SF is using the wrong vsync value when calculating the
target for ADPF, causing it to send a binder every frame instead of
every vsync change. This patch fixes that to reduce the binder
frequency from SF, and also adds a debug option to disable reportActual
to reduce the amount of binders even more in certain cases.

Bug: 284357218
Test: manual
Change-Id: I8435b170016c5af8219d8ed8cc9ed65a0a2cb0e7

Revert submission 23316821

Reason for revert: breaks touch on some devices b/284203485

Reverted changes: /q/submissionid:23316821
(cherry picked from https://googleplex-android-review.googlesource.com/q/commit:c96ed759fb428c7b982c921611e5da05040e3cea)
Merged-In: Id957e51b53e48b874302b24b612554f3024aa87b
Change-Id: Id957e51b53e48b874302b24b612554f3024aa87b

Revert submission 23316821

Reason for revert: breaks touch on some devices b/284203485

Reverted changes: /q/submissionid:23316821
(cherry picked from https://googleplex-android-review.googlesource.com/q/commit:342e1dd013c4c5f51a15222d5b112fb14583a636)
Merged-In: Ic4537a645d710a900d6a3b9328bd6489baedbb1d
Change-Id: Ic4537a645d710a900d6a3b9328bd6489baedbb1d

We're not aware of any condition that would lead this to happen. This
check is to help provide information in case we see future input windows
syncronization issues.

Bug: 279792237
Test: presubmits
Change-Id: I9bdcf34c71c1a741f863feb0881cd09122c7bfb2

Revert submission 23316821

Reason for revert: breaks touch on some devices b/284203485

Reverted changes: /q/submissionid:23316821

Change-Id: Id957e51b53e48b874302b24b612554f3024aa87b

Revert submission 23316821

Reason for revert: breaks touch on some devices b/284203485

Reverted changes: /q/submissionid:23316821

Change-Id: Ic4537a645d710a900d6a3b9328bd6489baedbb1d

This change fixes a bug where WindowInfosListenerInvoker's active
message count is incremented for messages when there are no listeners.
In this case, onWindowInfosReported is never called, leading to
out-of-sync window infos once listeners are added back. This scenario
may occur if system server dies.

Bug: 279792237
Test: WindowInfosListenerInvokerTest, manual tested by killing system server
Merged-In: If62f7cc56b48570f633e8e640006f020b053ea6f
Change-Id: I521dad46a763d0380807ba20b9a4ec9dcecf6bfc

This test will provide RawEvent's to TouchpadInputMapper and will look
at the returned events.

The main purpose of this CL is to document the existing behaviour of
CursorInputMapper and TouchpadInputMapper. The tests highlights that
today, dispatcher must handle the case where the touch starts without
HOVER_EXIT. That means that the dispatcher logs of "conflicting pointer
actions" can be mostly ignored.

The goal was to make the tests for the two mappers as similar to each
other as possible.

A slightly different testing infra is introduced here compared to the
one used in InputReader_test.

Changes:
* Use mocks for interfaces instead of constructed objects
This helps figure out which parts are important to mock for a specific
test vs which ones are not. When a function is called with a parameters
that the mocks aren't expecting, a warning is printed during test run.
This helps identify the complete state needed in order for the test to
execute.

* No longer require InstrumentedInputReader

* No longer require a listener. We only check the events that are coming
  from the 'process' call, which is what the interface for the mapper
  does.

Limitations:
* Still require an InputDevice object to be constructed in order to test
  InputMappers. Ideally, a mapper would only depend on the EventHub
  state (to read the current value of keys / axes after a reset).

Bug: 263319225
Test: m inputflinger_tests && $ANDROID_HOST_OUT/nativetest64/inputflinger_tests/inputflinger_tests --gtest_filter="*HoverAndLeftButtonPress*"
Change-Id: I7de0dee7abcf6bcb9d3283e29d9a85de2f331a44

Instead of using the displayId sent from WMS, it's better to use the
layerStack since it more accurately represents where the layer is
presented. Previously, the only case where WMS needed to set a displayId
different from layerStack was when using portal displays, which is no
longer supported.

This also fixes the case with cloned layers. If a cloned layer was
placed on a different display than the main layer, the WindowInfo for
the clone would end up having the same display as the main window since
the properties were directly copied over. With this fix, the cloned
layer's WindowInfo will get the correct layerStack that represents where
it's actually rendered.

Test: MediaProjection capture with mirror
Bug: 195280234
Bug: 194480991
Change-Id: Id1794798a543260c9868ccdc1be3cdb574b73dd5
Merged-In: Id1794798a543260c9868ccdc1be3cdb574b73dd5
(cherry picked from commit 6eada62e)

This reverts commit 0987fd88.

Test: m checkinput && atest libinput_tests inputflinger_tests
Bug: 256009910
Change-Id: I452b8292a5a797f21707f56ee01765650db99b94

Configure the Device prior to populating mappers for mappers to
receive correct property map

Test: m checkinput && atest libinput_tests inputflinger_tests
Bug: 256009910

Change-Id: I2a348029afa6c566506f1d79b655173bb8e7a8af

This partially reverts a change that triggered a performance regression
in InputFlinger.

Bug: 281227113
Bug: 281248006
Test: presubmits
Change-Id: Ia7bd668e3cab8a353b33f4fce4f98b6e57186b86

Screenshots can be taken for a partial hierarchy and may require
some of the layers to be temporarily reparented and their geometry
to be updated. Sceenshots use a copy of the layer snapshots and another
instance of the composition engine to render the contents to a
buffer.

There was an issue where the snapshots were updated before they were
copied and if the screenshot was taken during a period where there
were no visible region changes (and no reason to rebuild the snapshots)
we would pass invalid snapshot data when compositing to the display.

With legacy frontend, its harder to split the screenshot path and the
display composition path, so as a fix, update the snapshots after
screenshots are taken.

Bug: 282884552
Test: instrumented load to verify snapshots are updated correctly
Test: repro steps in bug

Change-Id: Ie778b02b57c5b1ddf0f09afe82675e04c5535ee5

Some tests force sRGB screenshots for consistent test output. Make sure
that output colorspace is still respected.

Bug: 283305205
Test: SurfaceFlinger_test
Change-Id: I42ffe36022fa65568bcd75df28a33389340a7d7f

Partially revert Ib114ffc1866aa32e6d493e06dcf5a27652d19a39.

PowerAdvisor needs the present fence of the previous frame, as the query
happens after presenting the current frame so should not be adjusted for
targeting 2 VSYNCs ahead.

Fixes: 283189729
Test: Perfetto
Change-Id: Ifbd6aa961bc92fe1d2f4de1247413fcd49330645

This change allows BackgroundExecutor to receive callbacks from multiple
producers.

Bug: 283133164
Test: BackgroundExecutorTest
Change-Id: I63f3266184a2ae90b15d7d1b7d9502847e268763

Add a factory method for KeyboardInputMapper to be configured on
initilisation

Test: m checkinput && atest libinput_tests inputflinger_tests
Bug: 256009910
(cherry picked from https://googleplex-android-review.googlesource.com/q/commit:67ca684ad1e5d6286bdbae568ee0fc97be40d3c1)
Merged-In: Iceb8676542de85309c208af93ee7cc385e46a067
Change-Id: Iceb8676542de85309c208af93ee7cc385e46a067