There were some integration tests in AutofillValueTest before, we
moved integration tests to CheckoutActivityTest. AutofillValueTest is
a unit test now, we don't need this test running under widget path.
Bug: 167847110
Test: TreeHugger
Change-Id: I514942cff0122539ac614798020ecbbaf7cbaa3c
formatStateDescription() in onProgressRefresh() is causing performance
degradation with accessibility off.
Note that stateDescription is set in onInitializeAccessibilityNodeInfo
to ensure that accessibility always get the update-to-date
stateDescription. Otherwise, stateDescription might be stale on first
accessibility focus and the user might get wrong information.
Fix: b/166215147
Test: tested with talkback enabled from the beginning or from the
middle.
Change-Id: I9fb99b479b172d87c8d0fa1b8765dab53445b993
- Remove the lineLeft/lineRight source bounds.
- Make the magnifier center X can be away from the cursor X.
- Let Magnifier know how to render floating cursor.
Test: locally verified.
Bug: 149883037
Change-Id: Iae3c68fa9cda0a83704fa30fa7b9f4db0a8d3314
Autofill broken tests (b/159037010, b/159035197, b/152444221) that
weren't executed in presubmit. Add the fail tests into TEST_MAPPING
to make sure the tests can be tested in presubmit.
Bug: 154961272
Test: atest --test-mapping frameworks/base/core/java/android/widget
Test: atest --test-mapping
frameworks/base/core/java/android/widget/inline
Test: TreeHugger
Change-Id: I718c86d8f55234a330235672016c877986c30d54
In Q, the slop is a fixed value calculated based on
TextView.getLineHeight().
In R (before this cl), the slop is a dynamic value based on the height
of the current line.
It's been confirmed as the root cause of the regression bug b/159528374.
This cl changes the slop value back to the fixed value.
Bug: 159528374
Test: Verified locally.
Change-Id: I182d22779f1c1155ae2e0b0deb0455bd270b2fd9
Merged-In: I182d22779f1c1155ae2e0b0deb0455bd270b2fd9
They can be replaced by the public API provided by ICU4J instead.
No extra layer of caching equivalent to the LocaleData cache is
added because ICU4J has caches instances, e.g. DateFormatSymbols,
etc.
Bug: 160606356
Test: atest FrameworksCoreTests:android.text.format
Change-Id: I07048e0e1a4835d2744b7fce6a5ed79a112e456b
In Q, the slop is a fixed value calculated based on
TextView.getLineHeight().
In R+ (before this cl), the slop is a dynamic value based on the height
of the current line.
It's been confirmed as the root cause of the regression bug b/159528374.
This cl changes the slop value back to the fixed value.
Bug: 159528374
Test: Verified locally.
Change-Id: I182d22779f1c1155ae2e0b0deb0455bd270b2fd9
The issue occurs because TextView#isFromPrimePointer assumes multi-touch
events are happening on the same TextView and relies on
ACTION_UP/ACTION_CANCEL to reset the Prime pointer id.
consider following scenario:
1. ACTION_DOWN (pointer 0) landed on ViewA,
ACTION_DOWN (pointer 1) landed on ViewB.
2. ACTION_MOVE landed on both ViewA and ViewB, triggering a drag.
3. When all fingers are released, ACTION_CANCEL (pointer 0) will be
landed on both ViewA and ViewB, so the prime pointer in ViewB is not
reset due to pointer index mismatch.
This CL always reset prime pointer id whenever ACTION_UP/ACTION_CANCEL
is triggered.
Bug: 157399040
Test: manual
Change-Id: Id35d9eac939378482fa50e85bb5cda58469240ad
Don't construct a new locale-specific NumberFormat every time a progress
bar's state is refreshed or updated. We now do this only when necessary
due to a locale change.
Bug: 161264248
Test: atest ProgressBarTest
Change-Id: I349ec28268a3471184e5746b164f81428ad80cc8
In RV, a view could be temporarily detached in one draw cycle.
Accessibility events sent during this time will be lost. To prevent the
events from being lost, we will set a flag when a view is detached by
parent. And when sending accessibility events for detached view, we will
delay it a little by
ViewConfiguration.getSendRecurringAccessibilityEventsInterval. This way,
if the view is attached in one draw cycle, the events will be sent out.
Also add more setStateDescription() in CompoundButton to for safer code.
(this is not the cause of this bug though).
Fix: 151125936
Test: tested that the bug is fixed.
Change-Id: Iffca8c87bad4fa2f66862b966e351562d77d6d76
