Although it may seems a left-over from a previous interrupted write,
actually there are callers who call startWrite(), openRead() and then
finishWrite(), and this was okay in the previous implementation, so we
have to keep supporting it.
The new file is virtually ignored in the new implementation, and we
have no good way to know if it's actually a left-over or one that's
being written, so simply leaving it there is also okay.
Fixes: 157092639
Test: atest AppIdleHistoryTests#testFilesCreation
Change-Id: I4dc7fde99d2b8e04356f082a6e6ad61c2835022e
Fix documentation to clearly indicate that the default behavior is to
show WebView's own default dialog, and also describing clearly default
dialog behavior and how to customize it.
Note that onJsBeforeUnload is not updated at the moment as I could not
find a way to reproduce it.
Bug: 154014645
Test: m -j offline-sdk-docs seems not broken
Change-Id: I5ee09ea35340eb8d17353eda1786dcebcff4a29e
Typically this data would have been flushed at the end of the
first draw and there would otherwise be an extended delay
before the data would be sent to the service.
Test: make -j
Test: Manually start WhatsApp - check for lost events
Test: atest CtsContentCaptureServiceTestCases
Bug: 154777879
Change-Id: Ia0a97e24ffcc1c189d5d0cd6b8888678fa48968e
This was moved to PackageCacher, but the old and unused counter
was not removed.
Bug: 154310064
Test: manual device reboots and logs cached count >0
Change-Id: I32fdb4b8fccd281fe61c64f231cb0ba154934679
A few months ago MediaProvider forked SQLiteQueryBuilder and we
fixed bugs in how the new insert(), update(), and delete() methods
were executed, but those fixes never made their way back upstream.
This CL is a clean cherry-pick of existing logic.
Bug: 155149941
Test: atest android.database.sqlite.cts.SQLiteQueryBuilderTest
Change-Id: I00d2ee88539b12d27a2ed99fff0546d1e3543b0e
Otherwise, we would think its visibility needs to be cleared.
This CL:
- reverts parts of eb5a5920b7,
- prevents sending redundant MSG_DISPATCH_SYSTEM_UI_VISIBILITY, and
- cancels clearing visibility if a new reversed visibility is set.
Fix: 156994866
Test: atest --iterations 100 LayoutTests#testAddingImmersiveWindow
atest WindowInsetsControllerTests InsetsAnimationControlImplTest
Change-Id: I8b8b7541523c0a94daf336f3be415aa5fa3f33da
The previous implementation of backing up beforehand doesn't handle
the case where the file is created for the first time, and might leave
a corrupted file in case of failure.
This new implementation creates a new file for writing data into, and
renames it into the place of the original file after writing
finished.
Fixes: 151959443
Test: atest android.util.AtomicFileTest
Change-Id: I5c4c438526a2aecdd2af18f71e16b41a05817c61
Merged-In: I5c4c438526a2aecdd2af18f71e16b41a05817c61
Update the deprecation note for WebSettings.setAppCacheEnabled and
related to mention the public blog post about the feature being removed.
Fixes: 156409857
Test: m offline-sdk-docs
Change-Id: I0481937d665f3f16b61b921aa19e306ba8bb16e0
The actual cache clearing is not async anymore, so we can report the
status to the calling app. Here, we just change the documentation to
make the caller aware of the 3rd possible activity result: EIO.
This value is returned if an error happens while MediaProvider is
clearing the cache. In this case, we don't know how much of the cache
was deleted, if at all.
Test: build
Bug: 155290762
Change-Id: I3a04a9b5b6b07c6ec61e97a4cb0d3bc8b10b0370
This change updates the docs for the MATCH_UNINSTALLED flag to note that
without the QUERY_ALL_PACKAGES permission, uninstalled packages will not
be returned.
Bug: 149846504
Test: builds
Change-Id: Id0c5b7f29172bd334dc1b2a32ce1f4eb7b1f0bd3
It requires a permission which we can't force apps to take to
maintain backwards compatibility. We also arguably cannot because
it leaks visibility, although only for debuggable apps/non-release
builds.
Instead, there's a new static method for getting the raw targetSdk
to gate against and the check is done manually, ignoring
enabled/disabled state. This will cause a mismatch between certain
apps and some system services like AppIntegrityManager, but the
effects should be minimal if we assume that most people ship
valid APKs. At worse the integrity check will pass an APK that
PM will fail, which doesn't break the feature.
Bug: 156356591
Bug: 156778241
Test: manual device boots
Change-Id: I877a5061476b86b9d63c34e75f16b38be8c3e1c2
