The fix in:
https://googleplex-android-review.googlesource.com/#/c/300346/
worked but the constant used had an extra trailing zero - which was confusing
and put a 1 in the 'flag' space of the measurement spec.
The intended number was:
0x00800000
Unfortunately, this intended constant doesn't fix this bug.
The constant submitted in this fix is:
0x00010000
which is outside the 'flag' space of measurement specs and appears to steer clear of overflow
problems in the scenario of this bug.
As suggested in the submission above, it would be preferable to rework of the RTL code to avoid
the use of such a constant as it seems very unlikely indeed that any choice of integer can
avoid problems in all cases.
Change-Id: I0c6744257ef2aebe8dbc8c041a447f9b90ee4b84
This patch makes segmentation into script runs behave the same in RTL
as in LTR modes - so that inherited script characters are always
associated with the preceding run. Otherwise, for a sequence such as
u+0631 u+064d u+0020, it would get split after the first character,
which would lose the ability to correctly position the u+064d mark.
Change-Id: I3c12ba1b77d18334f55e707f518be1046e6b339b
By adding these stubs, they will be automatically added to framework.aidl
in sdk builds.
This makes it easier for unbundled apps to pass these objects across
AIDL calls.
ContentValues & CursorWindow are already public Parcelables. It is an
oversight that they were not already in framework.aidl.
There are a lot of other public Parcelables that are missing from
framework.aidl. This just fixes two commonly requested ones.
Change-Id: If61e19b1206da1680413d9ea03de87a90b6d233e
VPN used to just watch the interface, but that is insufficient. There
is no promise that the interface will go down when we're done with it.
Now that wifi stays on in scan-only mode despite user turning it off
it seems that the interface is left up, even in AP mode.
Now listening for ConnectivityService broadcast that the network we were on
has disconnected and tearing down the VPN then or when the interface
goes away.
bug:8550083
Change-Id: Icf414497bc55bead69de04e91f39f90ac2e6578a
GridLayout is working as intended here. The bug is appears to be in RelativeLayout
(and possibly LinearLayout).
The value of RelativeLayout.DEFAULT_WIDTH = Integer.MAX_VALUE/2 is 0x3FFFFFFF has bits
set in the range that is used to flag certain conditions and states by the layout system.
In View we have:
MEASURED_SIZE_MASK = 0x00ffffff
MEASURED_STATE_MASK = 0xff000000;
MEASURED_STATE_TOO_SMALL = 0x01000000
This change fixes this bug, though it looks as if that a safer solution would be to not introduce
this constant and code path in the first place - as RelativeLayout's measurement algorithm operates
in the LTR case without it.
Change-Id: I01c51ae854620f08dd63047594486a3464c86f3a
The node id does not have to be decorated with spans like spannable
so it makes no sense to have these APIs use anything else but string.
bug:8657338
Change-Id: I2e7c31128ee9f2933bd0d58beac4ba31a498bb09
Protip: Don't mess with Bundles after you've sent them off
for parceling in an RPC.
Note that this change reduces the payload size of
StatusBarNotification objects received in
onNotificationRemoved() callbacks; it scrubs out the
RemoteViews and Bitmaps just as the NoMan's internal archive
does. [You don't really need that information anyway when
hearing about a removed notification; most likely all you
need are the other slots on StatusBarNotification, but
nulling the whole Notification object breaks a lot of
clients.]
Bug: 8616295
Change-Id: Ic899045f2352b96dcf064d3e9e51dad52629aea3
The current invalidation logic does not take into account the clipChildren
flag. When this flag is set to false on a container (an uncommon but
possible case), it is possible for views in the child hierarchy of
the container to be draw outside of the container's bounds. But invalidations
on that view hiearrchy can be clipped to the container's bounds, causing us to
not redraw views outside of those bounds.
Fix is to expand the dirty rect of an invalidation to encompass the complete
bounds of any container with clipChildren==false.
Issue #680037 Some transform combinations can leave old pixel values on the screen
Change-Id: I426beee15d04145fac2f6b4203748ae309e392b4
Fixed b/8276827
Vendor might want to provide their own implementation of "network
location", "fused location" and "geocoder" service. Location manager now
allows those service to be replaced by packages that have the same
signature as one of the packages in config_locationProviderPackageNames.
Such behavior might not be desirable on some devices. This change
make this behavior configurable by 3 boolean flags.
Details:
- Added three boolean flags in core/res/res/values/config.xml to enable
or disable NLP/FLP/Geocoder overlay
- Added 3 package name for the stock NLP/FLP/Geocoder. They are needed
only when overlay is disabled because LocationManagerService need to
know which package is preferred when searching for
NLP/FLP/Geocoder service.
- Made ServiceWatcher able to handle non-overlayable services.
- Fixed an NPE isue in ServiceWatcher. mPm.queryIntentServicesAsUser
might return null.
- Fixed an bug: justCheckThisPackage in bindBestPackageLocked is always
ignored.
Change-Id: Id221961ac7c3aa8ad44b894f9523f04f770ae237
The rendering code optimizes by rejecting drawing operations that
lie outside of the bounds of their views. This works in most
situations, but breaks down when containers have called
setClipChildren(false), because we reject drawing that is outside
of that container, but which should be drawn anyway.
Fix is to pass in the value of that flag to the DisplayList drawing
routines which take that flag into account when deciding whether
to quickReject any particular operation.
Issue #8659277 animation clipping
Change-Id: Ief568e4db01b533a97b3c5ea5ad777c03c0eea71
Only allow through changing direction configs for drawables.
Explicitly map layout direction values to an index in the
preload arrays.
Drawables that don't vary by configuration should go in to both
the rtl and ltr preloads.
Change-Id: Ib92dd11738082a795e02d1d4191adb54702d651c
SharedPreferences uses deferred writes internally, and the public
API doesn't allow apps to explicitly synchronize with this, so the
backup/restore implementation needs to take a little care to make
sure that the app process isn't killed before the deferred writes
land on disk. This parallels the coherence guarantees around
SharedPreference that the Activity and Service lifecycles provide.
Bug 8659368
Change-Id: I853e54f9fb0d2d260dbe6e40d640959f998092df
