We have three cases for handling AXIS_ORIENTATION:
1. Orientation is not supported by the input device, so the value for
   AXIS_ORIENTATION should always be 0, regardless of display rotation.
2. Orientation is supported, but a "direction" is not specified, like
   for touchscreens and touchpads. The orientation must be in the range
   [-pi/2, pi/2] for all display rotations.
3. Orientation is fully supported, and the value is in the range [-pi,
   pi] for all display rotations.

It is insufficient to rely on whether or not the PointerCoords has the
bit for AXIS_ORIENTATION set to determine whether the event has a valid
orientation. This is because we always skip setting values of 0 for any
axis in PointerCoords to save space during serialization.

To support these three cases, we introduce two new MotionEvent private
flags. These are flags that are not exposed to Java and to the public
APIs.

Bug: 263310669
Test: atest TouchScreenTest libinput_tests inputflinger_tests
Change-Id: Iaa38afe35b00de74fbc5eefce25191bea52c2ea6

This commit explicitly marks symbols in the code with default visibility
if they should be part of the libbinder ABI. Later, we'll set the
default symbol visibility to hidden and so any symbols not explicitly
marked as exported will be hidden.

We are still overexporting symbols a bit. When `LIBBINDER_EXPORTED` is
applied to a class, it exports all of that class's symbols. I've tried
to only export class members, but in a few cases it didn't seem worth
the trouble, e.g. where all the members need to be exported.

Additionally, we are exporting some symbols only needed by binder test
binaries. Since those tests run against the standard libbinder.so, we
need to include the symbols as part of the ABI.

On aarch64, this reduces the size of libbinder_on_trusty_mock.so by
10.7% and libbinder.so by 7.5%, primarily in the .text, .dynstr, and
.dynsym sections (but none of this takes effect until we actually
enabled it in a later commit).

Bug: 338458975
Test: TH
Change-Id: Idc348c2b12124a8bbcebeea49920267a995367b6

Optional<T>::ok_or maps to Expected<T, E> where nullopt becomes E.

FTL_TRY unwraps T for Expected<T, E> or does an early out on error.

Bug: 185536303
Test: ftl_test
Change-Id: Ia03f7e3d8773878db1c493b62772ab2c2b7a4fed

- Silence warning about signedness change.
- Disable copying to prevent stray end_ pointer.

Bug: 185536303
Test: Build
Change-Id: Ief31910bd51e9582a576ace554a5f7bafe46bf02

These will eventually replace the existing definitions as follows:
     ADISPLAY_ID_NONE -->> LogicalDisplayId::INVALID
     ADISPLAY_ID_DEFAULT -->> LogicalDisplayId::DEFAULT

We are keeping the old definitions for now, to reduce the merge
conflicts. These will be removed in subsequent CLs, after the
frameworks/base and other repository patches go in.

The constant "NONE" was renamed to "INVALID" to make it consistent with
the Java definitions.

Bug: 339106983
Test: m checkinput
Change-Id: I0274be345159c85cb51fcea743d8acd3d298cd07

The downstream branches now know about Parcel and binder, so in some
places, these #ifdef statements are not needed.

This CL should help reduce the divergence with downstream efforts. See
the merge conflict resolution CL as an example:
https://googleplex-android-review.git.corp.google.com/c/platform/frameworks/native/+/27201796/-2..3

Bug: 309829647
Test: presubmit
Change-Id: Iec17cface5069ca79a565c205546db8ada8a07ee

Currently, we use int32_t for displayId, which is not a safe type, and
it may also lead to misdefinition of types. Here, we introduce
LogicalDisplayId as a strong type for displayId and move all contents
of constants.h into LogicalDisplayId.h.

Bug: 339106983
Test: atest inputflinger_tests
Test: atest InputTests
Test: m checkinput
Test: m libsurfaceflinger_unittest
Test: presubmit

Change-Id: If44e56f69553d095af5adb59b595e4a852ab32ce
Signed-off-by: Linnan Li <lilinnan@xiaomi.corp-partner.google.com>

These functions intentionally overflows integers. The upcoming compiler
update, clang-r522817, will start complaining about these overflows.

Test: boot with new compiler
Bug: 325934863
Change-Id: Idb3730c652ae739b73c09517665b518a79b60345

The surface control and transaction are barely every null in most calls,
but some informational structures like HDR metadata, and the surface
control when reparenting are allowed to be null.

Change-Id: Ia1a456041daf8136d856a1901881b74f97b3593a
Merged-In: Ia1a456041daf8136d856a1901881b74f97b3593a
Test: make aosp_cf_x86_64_phone
Signed-off-by: Marijn Suijten <marijns95@gmail.com>

The messages come from a different CL in AOSP, but that CL breaks
internally because this code doesn't exist in AOSP. Merging the
compatibility thing first so I can land the rest in AOSP.

Bug: None
Test: None
Change-Id: I906c6ea7bd6d891aba9e185756131ac5b34f3c3d

"Unrecoverable internal errors" here means that ALOOPER_POLL_ERROR
will be returned if the internal epoll_wait call fails with something
other than EINTR. I could say that explicitly, but I think the
important info for the caller is "you're doomed, just abort".

Bug: None
Test: None
Change-Id: I869f29122500ee8c71390cae1c28108a63088a77

This (hopefully) explains all the bits that confused me when I tried
to use these APIs for the first time. The existing docs leaned a lot
on developers already understanding the Java API, and even the Java
API docs didn't explain how to use this portion of the API well.

Bug: None
Test: I am the test
Change-Id: I48490112f92ef754b38daef7a4ebf6c031cc20f8

They're deprecated, but still needed by most users, as the replacement
is not available until API 29. Restore the docs and explain the bugs.

Bug: None
Test: treehugger
Change-Id: I4227ce3f0ee57a73a13e394cd48373accf4a58a3

Add the soon-to-be-required removal message for ALooper_pollAll, and
expand the docs for both ALooper_pollAll and ALooper_pollOnce to make
the reason for removal clear, and to explain the fix callers should
apply.

Bug: https://github.com/android/ndk/issues/2014
Test: treehugger
Change-Id: I020a280346d1ff9f5fcbec4c153dee8319cdf539

Bug: None
Test: treehugger
Change-Id: I3a52883fba0e31f05e50cd51193bbb306418b18a

* Disallow NonNull<bool> by checking for nullptr, instead of false.
* Add an "operator bool()" to work like other smart pointers.
* Add [[nodiscard]] where it makes sense.
* Implement the full set of comparison operations, and allow
  comparisons between NonNull<P> and another type Q. This allows the
  use of NonNull<P> in sorted containers like std::set, and allows
  easier lookup via unwrapped pointers.
* Specialize std::hash for NonNull. This allows the use of NonNull<P> in
  hashed containers like std::unordered_set.

Test: atest ftl_test
Bug: 185536303
Change-Id: Ib1090e393ea5f5641be2cd9c61011049039ea615

ftl::contains(container, value) returns true if the container contains
the value. It's implemented in terms of std::find(), so has the same
runtime complexity.

It is otherwise a simplified version of the C++23 std::ranges::contains,
which has some additional options, and can be deprecated once C++23 is
the minimal supported version.

Test: atest ftl_test
Bug: 185536303
Change-Id: I98aefe7cf6645ac3a20fddfe0657fa6822d669de

This is already useful in several places, and it's going to be also
useful in the future when InputConsumer is refactored.

So let's move it to include/input.

Bug: 311142655
Test: m inputflinger_blocking_queue_fuzzer
(cherry picked from https://googleplex-android-review.googlesource.com/q/commit:3782af62ea590e8945f2175a90aa2b6f4995814d)
Merged-In: Idfc492c6bfc3cccab7e0b0d12b21a41a954cc44b

Change-Id: Ie662961b77d331cd62b529bae0fd66e56a69bd19

To ensure that the future updates are consistent across java, cpp, and
rust, we move all of motion event flags to aidl. That will be the
source of truth for those constants. That still doesn't ensure that in
the future updates, we forget to add these to all the languages, but at
least now it will be much easier to do so.

Bug: 309829647
Test: m sync
Change-Id: Ibb92c0c2b2da3b07eac3886c6fd652a295037099

Extends stable hash from libui by increasing maximum string length
supported from 16 to 64 bytes (CityHash64 v1.0.1 - please see
http://code.google.com/p/cityhash/).

Bug: 185536303
Bug: 194863377
Test: ftl_test
Test: Compare with 64-bit std::hash<std::string_view>
Change-Id: Ib123d0690a0d2150aa51c91526784dcf77a65d46

These occur when an enum value is formatted as an integer without a
cast. In all these cases the debug logs are nicer if we use the actual
name of the enum value, so wrap them all in ftl::enum_string calls.

Bug: 245989146
Test: Check there are no -Wformat warnings in the output of
      $ m checkinput
Change-Id: I26333123e9fb445f1640c5e73757ddf4ad2d6e98

Adds a hidden method for the creation of special "internal" hint
sessions with extra metadata, and plumbs hwui to use it

Bug: 330553312
Test: atest PerformanceHintNativeTestCases
Test: hwui unit tests
Change-Id: I2e7eb8fe010b7314c3e91ec2d11ac1910c234262

We currently display the pointer for all devices that support displaying
it, but there are actually some devices that are disabled and shouldn't
actually have it displayed for them, here's an implementation for that.

This is done by adding an enabled state to the InputDeviceInfo to
determine its disabled status.

Bug: 335795596
Test: atest inputflinger_tests

Signed-off-by: Linnan Li <lilinnan@xiaomi.corp-partner.google.com>
(cherry picked from https://partner-android-review.googlesource.com/q/commit:d7c5c6d376f5eae7cbb7cce5ea12f2ba420d87ab)
Merged-In: Ibdad9f8424a29d83412866eace0ef96026fb270a
Change-Id: Ibdad9f8424a29d83412866eace0ef96026fb270a

The surface control and transaction are barely every null in most calls,
but some informational structures like HDR metadata, and the surface
control when reparenting are allowed to be null.

Change-Id: Ia1a456041daf8136d856a1901881b74f97b3593a
Test: make aosp_cf_x86_64_phone
Signed-off-by: Marijn Suijten <marijns95@gmail.com>

Change-Id: I6e18c9d99d562aec112fc8ecb3d4e916285368c6
Test: Built

This keeps it consistent with the SDK doc.

Bug: 332253611
Test: n/a
Change-Id: I988f9be7eb37ac1d57edd403dc30f0e17e2b2db3

By declaring the ABI of `enum` types, the `enum` type can now be used in
place of a raw and untyped integer in function parameters and structure
fields, self-documenting what the possible values are.

Test: treehugger
Change-Id: I6b9cf8b36b16e23c0e992f8155452bd194e6e1ad
Signed-off-by: Marijn Suijten <marijns95@gmail.com>

Bug: https://github.com/android/ndk/issues/1920
Test: treehugger
Change-Id: Ib8296e7b9c86809d16fbde0787d016095a189387

This will change MotionPredictor::predict() outputs once
enable_prediction_pruning_via_jerk_thresholding is enabled.

Test: atest libinput_tests --host with local flag override to enable
enable_prediction_pruning_via_jerk_thresholding.
Test: atest CtsInputTestCases
Test: atest MotionPredictorBenchmark MotionPredictorTest
Bug: 266747654

Change-Id: I11eb1972246468a1f3824656f5ac57e01e0359cd

This will help understand when it's created and destroyed. This also helps disambiguate between multiple InputConsumers in the same process.

Test procedure:
1. Add "system tracing" to quick settings from developer options (enable
   the quick settings tile for system tracing)
2. Start system tracing via UI of quick settings
3. Open all apps and do some swipes
4. Click "stop tracing"
5. Pull the logs to the device `adb pull /data/local/traces/ ~/trace`
6. Go to https://ui.perfetto.dev/
7. Click "open trace file" and navigate to the local file
8. In the trace, open the process of interest. In this case, it's the
   launcher process
9. Scroll down to "InputConsumer processing"
10. The "cookie" field of each entry is the sequence number of this
    event (but not the event id).

Bug: 311142655
Test: perfetto test procedure described above

Change-Id: If8cdadd3604e8e3ab24c45a5ccf4ea9ffa9fee4a

This is a no-op

Test: atest libinput_tests
Bug: 266747654
Change-Id: I0f67a1d8ea59a6851afa6f523b6e00b2281745d0

to be shared between Java and C++.
The values of PointerIconStyle in Input.h and the constants in
PointerIcon.java are manually synced now.
This change introduces an AIDL file which can be used by both.

Bug: 235023317, 330827347
Test: presubmit
Change-Id: I01f31868e7c13270047673d97035783deaf0a7f6

Because of the existence of extra spaces, the owner of keycodes.h is
not configured as expected. We should delete this extra spaces.

Bug: none
Test: After merging, check whether the owner of keycodes.h includes
all users under INPUT_OWNER

Change-Id: I634c49f53a4633bdee3cdefbf0debae4dff90a47

Test: Builds
Bug: 330375394
Change-Id: I54a51e3f78d5c44067369d3fce3a61fd18a4157b

Bug: 324271765
Test: ASurfaceControlInputReceiverTest
Change-Id: I19a1c796c1d22e8b247368df07c34c1b4b195e64

This will allow the upper layers to release the ownership and hand the
object over to the ndk client. The client will then assume the
ownership. This is not possible with the current implementation of
rvalue, because the ndk interface is only using pointers.

Unfortunately, this makes things more clunky for tests (unless we modify
BlockingQueue).

Bug: 324271765
Test: TEST=libinput_tests; m $TEST && $ANDROID_HOST_OUT/nativetest64/$TEST/$TEST --gtest_repeat=100 --gtest_break_on_failure
Change-Id: I0cf326a22f840be6f8aa00d1e69f818815788487

In this CL, an InputConsumer with built-in looper handling is being
added. This will be useful for native code that needs batching, and that
provides choreographer callbacks directly into native code.

Before this CL (and with this CL, temporarily) the InputConsumer logic
was split between native and jni layers. In general, we shouldn't have
logic inside jni. But in this case, the situation was also making the
code difficult to reason and debug.

In this new InputConsumerNoResampling class, all of the features of
InputConsumer and NativeInputReceiver are combined, except for
resampling. That will be done separately, at a later time.

As a result, we will not be switching to the new
InputConsumerNoResampling class right away.

Once resampling is added, we can switch to the new
InputConsumerNoResampling (and rename it to InputConsumer), and delete
the old InputConsumer.

In the meantime, the new InputConsumerNoResampling will be useful in the
new NDK APIs. There, having resampling is not critical.

Bug: 311142655
Test: TEST=libinput_tests; m $TEST && $ANDROID_HOST_OUT/nativetest64/$TEST/$TEST --gtest_repeat=100 --gtest_break_on_failure
Change-Id: I468ddbd8406c4bf9f5e022f79fd1a582ba680633

Test: Builds
Bug: 203080148
Change-Id: I586339b82e408e0ac90944ab280a2d5b79309141

This helps to keep track of things, and it also allows easier deletion
in the future.

Bug: 311142655
Test: none
Change-Id: Iece2bf6857ab05f86072031a4cf3a36f843b1634

Added a native class that corresponds to the Java InputTransferToken.

Test: SurfaceControlInputReceiverTests
Test: AInputTransferTokenTest
Bug: 324271765
Change-Id: Ida2a7b34338560dfed9af7f510d64372e41384af