| Age | Commit message (Collapse) | Author |
|---|
|
This splits the monolithic 'tests' target into multiple types
based on classification. See the corresponding README.md for
descriptions.
The target 'IntentResolverUnitTests' is replaced by:
IntentResolver-tests-unit
IntentResolver-tests-activity
IntentResolver-tests-integration
IntentResolver-tests-shared
To run everything, use:
atest IntentResolver-tests-*
To run TreeHugger presubmits:
atest --test-mapping pacakges/modules/IntentResolver
Equivalently for postsubmit:
atest --test-mapping pacakges/modules/IntentResolver:postsubmit
Bug: 300157408
Test: atest IntentResolver-tests-*
Change-Id: I9d499284f070a4bfa3e7c1b7c3bbfaa8adb3379b
|
|
|
|
Each interface has a single concern, allowing for list controllers to be
built by composition.
New list controllers are not currently in use, but will be in a future
change once resolver comparators and list adapters get updated.
Test: atest com.android.intentresolver.v2.listcontroller
BUG: 302113519
Change-Id: Ie1d24571c07d1408aa80f8a86311d0fee5e78255
|
|
Adds requestState to allow modification of user profile state,
including availability (quiet mode).
Test: UserRepositoryImplTest
Change-Id: Ic38f24475c73390841ee599c48d965117981faa0
|
|
ActivityLogic" into main
|
|
Test: atest com.android.intentresolver.v2
BUG: 302113519
Change-Id: I590c10264220e5d328cad057215e5c4d8b120a4a
|
|
```
import androidx.core.content.getSystemService
val userManager: UserManager = lazy { context.systemService()!! }
```
Types are inferred as expected, (returns T?)
Change-Id: I7ad3531397d6afea25f25bb27c9f29d31a470ca2
|
|
Automated cleanups:
use androidx.annotation.* consistently across the codebase.
corrects missing nullability annotations method parameters,
constructors and and overridden methods return values.
Test: atest IntentResolverUnitTests
Bug: 300157408
Change-Id: Id48b7ce3e70400bd8ff5d51dabc77e8e04bbfc5c
|
|
An abstraction of users and profiles, packaged up into an injectable interface.
UserDataSource
val users: Flow<Map<UserHandle, User>>
fun isAvailable(@UserIdInt userId: Int): Flow<Boolean>
Along with an interface and implementation this change introduces a data model,
[User] to abstract from platform internal types.
Bug: 309960444
Test: atest FakeUserManagerTest UserDataSourceImplTest
Change-Id: I46681e5f5b40c0720f4b99c1bb13d05ab5da4211
|
|
* changes:
Moves AnnotatedUserHandles to ActivityLogic
Moves theme and profile switch message to ActivityLogic
|
|
Also creates test activity logic for wrapper activities to be compatible
with the ActivityLogic structure.
Test: atest com.android.intentresolver.v2
BUG: 302113519
Change-Id: If064e76015e90fa98d9e41fb3c8f38b93ccae789
|
|
Allows configuring a mock with different default settings
while still using a fluent Kotlin friendly syntax.
Also adds THROWS_EXCEPTION, which can be used as a default
answer within MockSettings:
mock<Context>(withSettings().defaultAnswer(THROWS_EXCEPTION)
Change-Id: I8aadc5da5aaaff865a7a163dd76ea047b61fd2d6
|
|
In content preview classes, Lifecycele was used as coroutine provider.
This change makes CoroutineScope as a depenedency for all those classes.
Test: atest IntentResolverUnitTests
Change-Id: I2d9f07a54d9bb5b56b20c19a821e4626c86da472
|
|
The tests sill remains undeterministic (we'd need some structural
changes in the app before we'd be able to change that) so the change
just increases timeouts.
launchActivityWithTimeout method is removed, the test timeout is used
instead as an equal but more simple solution.
Bug: 304822572
Test: atest IntentResolverUnitTests
Change-Id: Ibd91f1c02f17fe49040182dd589bc9abb3d5308f
|
|
|
|
Bug: 285314844
Test: presubmits, funtionalityf smoke tests
Change-Id: Ic2226af153850c6855c08ca20063fcc65c8b958e
|
|
Bug: 255714762
Test: presubmit
Change-Id: I98872680c2294353022f09ea5cd01cc048f04e12
|
|
|
|
Adds a SecureSettings fake
Adds test coverage for new modules
Removes another overload from ChooserActivityWrapper
Uses @BindValue in tests to alter the configured editor component
Test: atest --test-mapping packages/modules/IntentResolver
Bug: 300157408
Bug: 302113519
Change-Id: Ie7d5fe12ad0d8e7fd074154641de35fe89d50ce6
|
|
Fix: 297551329
Test: atest IntentResolverUnitTests:ChooserActionFactoryTest
Change-Id: I0241f840a3add9c433b312f440d5b12e9129ad30
|
|
As part of ongoing work to pull low-level empty state responsibilities
out of the pager adapter, this CL generally replicates the change
prototyped in ag/24516421/5..7, in smaller incremental steps to "show
the work." As originally described in that CL:
"... moves most of the low-level logic of `showEmptyState()` into
the UI helper. (...) also has the UI helper take responsibility for
setting the visibility of the main "list view" in sync (opposite of)
the empty state visibility."
As presented in this CL, the "incremental steps" per-snapshot are:
1. Extract most of the implementation directly to the new method at
`EmptyStateUiHelper.showEmptyState()`. The general functionality
is covered by existing integration tests (e.g., commenting-out the
new method body causes `UnbundledChooserActivityWorkProfileTest`
to fail). New `EmptyStateUiHelper` unit tests cover finer points
of the empty-state "button" conditions, and I've added a TODO
comment at one place legacy behavior seemingly may not align with
the original developer intent.
2. Also make the UI helper responsible for propagating empty-state
visibility changes back to the main list view (hiding the main
list when we show an empty state, and restoring it when the empty
state is hidden).
3. Look up all the sub-views during `EmptyStateUiHelper` construction
so we don't have to keep repeating their View IDs throughout.
4. Tighten visibility on `EmptyStateUiHelper.resetViewVisibilities()`
now that it's a private `showEmptyState()` implementation detail
(updated to package-protected/visible-for-testing). Also move the
method to the end of the class (after all the public methods).
Bug: 302311217
Test: IntentResolverUnitTests
Change-Id: Iac0cf3d62e2c3bf22afa6a2796ae4e731b706c02
|
|
As part of ongoing work to pull low-level empty state responsibilities
out of the pager adapter, this CL generally replicates the change
prototyped in ag/24516421/4..5, in smaller incremental steps to "show
the work." As originally described in that CL:
"... moves ownership/usage of the 'container bottom padding
override supplier' over to the UI helper; this PagerAdapter dependency
is really only used in empty-state logic, so it can be handed over to
the extracted component *instead of* retaining it as an ivar in the
PagerAdapter."
As presented in this CL, the "incremental steps" per-snapshot are:
1. Move `getActiveEmptyStateView()` from `ChooserActivity` to
`MultiProfilePagerAdapter`, and expose it in favor of the broader
`getEmptyStateView()` method that `ChooserActivity` previously
used to implement the lower-level logic.
2. Remove `MultiProfilePagerAdapter.setupContainerPadding()`
parameters since they're just derived from a combination of a
single hard-coded (resource) symbol, and a concern that already
belonged to the pager-adapter (`getActiveEmptyStateView()`).
Switch tests to using one of our real layouts so that we can find
an empty-state container with the expected View ID.
Also update method visibility to show `setupContainerPadding()`
isn't intended to be overridden in the more-modern "generic pager"
design, and update Javadoc accordingly.
3. Inline `ChooserMultiProfilePagerAdapter.setupContainerPadding()`
as part of its `setEmptyStateBottomOffset()`, since the latter
method is never called anywhere else (and both concerns already
belonged to the specialized pager-adapter).
4. Refactor the internal `setupContainerPadding()` to operate on
profile descriptors instead of container Views. This is based on
the observation that both callers of `setupContainerPadding()`
were passing containers that they had to query from the
descriptor's empty-state view anyways, and the lookup used the
same View ID in both cases. This change encapsulates those shared
concerns at a higher level of abstraction.
5. Move the implementation logic of `setupContainerPadding()` into
the descriptor inner-class; given the override-supplier, the rest
of the operation can be implemented just using the info available
to the descriptor.
6. Move ownership of the override-supplier into the descriptor (to be
shared among all descriptor instances). The outer pager-adapter
had no remaining need to reference the supplier, and we can easily
confirm that this results in `setupContainerPadding()` calls for
each descriptor instance using the same supplier instance they
would've used before.
7. Moves implementation logic (including [shared] ownership of the
override-supplier) from the descriptor down to the empty-state
helper which now encapsulates the entire `setupContainerPadding()`
operation with zero args.
This may require *slightly* more of a leap-of-faith than the other
"steps," but note the extensive test coverage: the existing
`MultiProfilePagerAdapterTest` covers this exact functionality
as integrated into a broader config; parallel unit tests are
newly-added in `EmptyStateUiHelper`; and even our "broad-scope"
integration tests exercise empty-state logic to _some_ degree.
8. Inlines some client usages to simplify scaffolding that we don't
really need after the earlier refactoring changes.
Bug: 302311217
Test: IntentResolverUnitTests
Change-Id: Icc460ede5b8a0314d5c807dd884e6e2b7044bee9
|
|
Just cutting over to our new experiment infrastructure before I
continue porting over the changes that were originally prototyped in
ag/24516421.
Bug: 302311217
Test: IntentResolverUnitTests
Change-Id: Ied1843bee2be6ffb42ba4f539f6168a9d07a77d9
|
|
The forked versions are flag guarded by the ChooserSelector.
Test: atest com.android.intentresolver
Test: adb shell pm resolve-activity -a android.intent.action.CHOOSER
Test: Observe that the action resolves to .ChooserActivity
Test: adb shell device_config put intentresolver \
com.android.intentresolver.flags.modular_framework true
Test: Reboot device
Test: adb shell pm resolve-activity -a android.intent.action.CHOOSER
Test: Observe that the action resolves to .v2.ChooserActivity
BUG: 302113519
Change-Id: I59584ed4649fca754826b17055a41be45a32f326
|
|
|
|
Based on prototype work in ag/24516421, this CL introduces a new
component to handle the empty-state UI implementation details that had
previously been implemented in `MultiProfilePagerAdapter`, since those
details significantly clutter the implementation of that adapter's
other responsibilities.
As in ag/24516421 patchset #4, this just sets up the boilerplate and
kicks off with some "low-hanging-fruit" operations. Follow-up CLs will
continue migrating these responsibilities as in ag/24516421 (except
with more incremental testing).
Bug: 302311217
Test: IntentResolverUnitTests, CtsSharesheetDeviceTest
Change-Id: Ie9bb7f4e97836321521c3cf13c77cafc97b1a461
|
|
Add ScopedAppTargetListCallback class to wrap AppPredictor callbacks
that breaks the callback reference chain when the specified scope gets
closed.
Bug: 290971946
Test: Launch Chooser multiple times in a row, collect heapdump with
a forced garbage collection, verify that:
* there are no ChooserActivity objects in the dump;
* there are multiple ScopedAppTargetListCallback instances in the dump
with AppPredictor$CallbackWrapper in their GC root path.
Test: manual functionality smoke tests: activity configuration changes,
pinning targets (all triggers content reloading).
Test: unit tests
Change-Id: I5099eb7527098a90b3e00bb848eb41e2bc7d14d6
|
|
into main
|
|
|
|
Plus some trivial changes to improve code readability
Test: atest IntentResolverUnitTests:ResolverListAdapterTest
Test: atest IntentResolverUnitTests:ChooserListAdapterDataTest
Change-Id: Id32460fb3ea1d8a706e48361c9baa8cce67fd46f
|
|
This CL just contains the initial code-moves to group relevant classes
(including some inner classes/interfaces that are being pulled out of
`MultiProfilePagerAdapter`), without any other anticipated fixes or
reorganization. This corresponds to snapshots 1-3 of the prototype
ag/24516421 (to be followed with the other work that was started in
that CL).
Bug: 302311217
Test: IntentResolverUnitTests, CtsSharesheetDeviceTest
Change-Id: If82a761193f6ff0605c5a46c106f7c95699350c0
|
|
|
|
|
|
That is, this CL consolidates `GenericMultiProfilePagerAdapter` up
into the base `AbstractMultiProfilePagerAdapter` (and renames to drop
the prefixes). This advances an in-progress refactoring to simplify
these classes (as in an earlier `GenericMultiProfilePagerAdapter`
TODO comment when that class was first introduced).
This prepares for other upcoming work in the base class (especially in
service of "private space" support) since the responsibilities were
unclear and arbitrarily split between these two bases, which had no
other direct clients/subclasses (i.e., this CL is exactly equivalent
to the earlier code, but removes a class from the hierarchy).
I've also started a little bit of cleanup around method visibility but
left much of that work out-of-scope for now.
Test: IntentResolverUnitTests / CtsSharesheetDeviceTest
Change-Id: I073e0bec5764ea16736af585af0bc9f744089d03
|
|
With the new `mDestroyed` bookkeeping in ag/24854314, we no longer
need the richer (but less-testable/etc) API of a `Handler`.
In the first snapshot, this retains the ad-hoc flow control mechanisms
that direct the `testPostListReadyAtEndOfRebuild_` tests to show the
direct before-and-after equivalence against the `Handler` model. All
the existing tests still pass on the new code.
The second snapshot removes those ad-hoc mechanisms because the tests
now have complete control of the execution via `TestExecutor`.
Bug: (general code cleanup)
Test: `IntentResolverUnitTests` / `CtsSharesheetDeviceTest`
Change-Id: I3063d45aedec47a81bf9bc7f76c26873de1383af
|
|
|
|
This change is best understood through the newly-introduced
`testPostListReadyAtEndOfRebuild_` tests:
1. The `synchronous` and `stages` variants pass both before and
after this change, to show the expected "normal" behavior.
2. The `queued` variant exercises the hypothetical bug that motivated
this change; it only passes once the legacy "throttling" condition
is removed in this CL.
3. The `skippedIfStillQueuedOnDestroy` variant covers a side-effect
of the fix. The original implementation had logic for this
"skipping" condition (and presumably would've passed this new
test as-written), but because the fix relaxes the invariant where
we can have only a single "active" callback instance at a time, I
also had to introduce the new `mDestroyed` condition to effect the
same high-level "cancellation" behavior. (This also addresses a
minor theoretical bug where the "destroy" could race against our
internal two-stage asynchronous flow, such that we'd end up
posting the follow-up callback to the handler after we'd
supposedly already been destroyed. I didn't think it was important
to test for this bug separately from the other coverage of the
`mDestroyed` condition.)
--
This CL removes a "throttling" condition that was unneeded, but which
could cause a hypothetical bug (reproducible in tests). The original
condition prevented list-ready messages if we'd already had one posted
in the queue (but not yet processed); however, there was no check
that the messages had the same parameters to indicate "partial" vs.
"complete" progress. Since the legacy mechanism favored the earlier
messages, we could end up dropping the one-off "completion" message
where our listener actually does work -- i.e., if we had to drop
messages, this is exactly the one we would want *not* to drop.
I believe this "throttling" mechanism was likely an optimization to
support the legacy `ChooserTargetService` design; the simplified
newer design requests fewer callbacks and so the throttling should
rarely come into play (but presents a bigger risk whenever it might).
Even in the older design, I suspect there would've been a risk of the
same "dropped completion" bug.
And, of course, it's nice to "simplify" by removing this condition,
even if it *weren't* strictly harmful.
Update: the second snapshot removes the old "callback removal" on
destroy, since the new `mDestroyed` condition gives effectively the
same behavior. Technically that's the only reason we depended on
`Handler` and we could now switch to using an `Executor` or etc -- but
I definitely want to keep that update to a separate CL.
Bug: as described above
Test: IntentResolverUnitTests, CtsSharesheetDeviceTest
Change-Id: Ifda9dc9a8ac8512d241e15fe52f24c3dea5bd9e7
|
|
|
|
This only covers the most salient functionality since many of the
implementation details are already in line to be extracted to separate
components anyways. The tests are being provided in advance of a
refactoring CL to merge this up into the ("Abstract" pager) base in
ag/24512960.
Bug: 286249609
Test: (this is the test)
Change-Id: I753e2d80639865a3c66a94b5e0a08e4be88654b3
|
|
|
|
|
|
|
|
Base feature implementation controlled by a flag. A few issues are known
and will be addressed separately.
Known issues:
* No (dis)appearance animation for the A-Z targets divider bar.
Bug: 287102904
Test: Enabled the flag programmatically and test the new functionality.
Manually test for possible regressions with the unset flag.
Test: atest com.android.intentresolver.contentpreview
Test: atest IntentResolverUnitTests:UnbundledChooserActivityTest
Change-Id: I8273cf365a1e00b1acff4030086f1a044ad7531f
|
|
DisplayResolveInfo#getDisplayLabel contains a logic to load the label
if it is missing and the majority of app targets in Chooser are loaded
this way.
This CL:
* removes label loading logic from DisplayResolveInfo;
* adds explicit label loading with TargetDataLoader, where it's needed;
* wrap some of the view related code blocks from
ChooserListAdapter#onBindView into ViewHoder methods to reduce
branching in the method and make it easier to read.
The legacy DisplayResolveInfo lable-loading logic is effectively
replaced with LoadLabelTaks's logic, the one that is used by resolver.
Bug: 289264582
Test: manual testing: labes loading, targets groupping, targets pinning.
Change-Id: I86814b5a4c67bf117fb1ea28c1d9980b5cf28ef5
|
|
Replace AsyncTask usage with an background executor and posting on a
main thread hanler to facilitate testing.
Tests are mostly around targets resolution logic in the adapter.
Test: atest IntentResolverUnitTests:ResolverListAdapterTest
Change-Id: I7af047226aa718ca3052aa4284d1e9d2a4c43ded
|
|
"Sticky" headline is one of the requirements of the scrollable preview
feature. Currently headling row is included in every preview type we
have and to be "sticky" it should be moved out of all of them. This
change is a preparation work that makes all of the preview UI
controllers aware of a possible external headline row (but no actual
external headline view is used).
Bug: 287102904
Test: manual testing of all preivew types checking that there are no
regressions
Test: atest com.android.intentresolver.contentpreview
Change-Id: Ib6110aaa82246e3354fdce18f431ab1c116e7b73
|
|
|
|
|
|
* changes:
Switches to HiltTestApplication in tests.
Extract an interface for EventLog, rename implementation
|
|
Using Hilt we are able to declaratively replace modules
within integration tests. This is the first such instance
within IntentResolver.
Scope: An InstanceIdSequence in @Singleton scope provides
a new InstanceId value to each EventLog created. EventLog
is @ActivityScoped or one-per-activity instance. This matches
the existing behavior.
By adding [TestEventLogModule], all integration tests are now
using a FakeEventLog by default (migrated away from 'override
data' scheme).
Bug: 299610743
Bug: 300157408
Test: atest IntentResolverUnitTests
Change-Id: I33d6f4d1241a890ab88b631859652117ce20f7be
|