Introduces new "net_bw_stats" group which will protect reading
detailed bandwidth statistics from the kernel. Also introduce
"net_bw_acct" group which will enable specific applications to
request that their network traffic be counted against other UIDs.
This change associates manifest permissions with the low-level GIDs.
Change-Id: If3fc28053afda201ff305d798a8878ff1f35b997
When SurfaceTexture is in sync mode, the consumer needs to pull the pending
frames. Otherwise the onFrameAvailable won't be called and no more update.
This simple fix is to skip frame to keep up with the producer side.
Change-Id: Ica833695dc0b0221d2543f46f0dd41aa55d1aa16
Previously we only sent hover to the topmost child, but this doesn't
handle cases where multiple children are overlapped to achieve
certain special effects. Now we send hover to all children until
one of them handles it.
Also moved the call to send the accessibility event into the
main dispatch function so that we can send the accessibility event
for all innermost hovered views even when setHovered() might
not be called.
Change-Id: I6fb8b974db44b594c441deafc012b8415afdfac7
- revert previous fix (transforming >=0 to >0)
- initialize mUserPaddingStart and mUserPaddingEnd to -1 in the specific View(Context context) constructor
- cache mUserPaddingStart and mUserPaddingEnd a bit earlier in View(Context context, AttributeSet attrs, int defStyle) constructor
- cache also mUserPaddingStart and mUserPaddingEnd in setPaddingRelative()
Change-Id: Ibe398da7af35f77655d7006444fa49c9c5387f62
Modified onHoverEvent so that a view only has its state changed to
hovered when the view is clickable. The intent is to prevent labels
from taking hover focus away from their containing view group despite
not being actionable in any way. The child will still receive
a hover event but, if it does not handle the event (returns false),
the hover event will bubble up to its parent and ancestors allowing
them a chance to handle the event instead.
The new onHoverEvent semantics are better because now they closely
mirror those of onTouchEvent. This makes it straightforward to
implement views that respond to hover by changing their visual
appearance (such as by making buttons glow when hovered).
Added onInterceptHoverEvent to enable ViewGroups to explicitly
intercept hover events within their bounds.
Exposed the new hover event API for real.
Change-Id: I63195c8f5c74b859f6047487f9a0f703e8f40ffe
Tainted events should still be processed, even if we don't bother
the developer with the details.
The check for nested calls to the event verifier must incorporate
the event type because different event types might be overridden
in different subclasses so the nesting depths may vary.
Change-Id: If2fb213e07db783a0bd1da97bb12aad93126d712
1. The query string was not converted to lower case during search
for infos with text.
bug:4818098
Change-Id: Ie4ba8ac2632072bdd71070ec8fd24ca255560ac0
1. While mapping Android hover to WebKit mouse events I was not
taking into account the horizontal and vertical scroll.
bug:4807087
Change-Id: Ie24a8a30ce6a20d27b627ae7c7fbff1b74d140c8
1. Updated the code to return empty list instead.
2. Updated the doc to explain that the search is done in the
sub-tree rooted at the info whose method is called.
Change-Id: I560d2563ed4772305bf2d7919560ef19baa30a1d
This adds a Java-language implementation of EventRecurrence.parse(),
to make it easier to relocate it for the benefit of unbundled
Calendar.
Differences from the native version:
- enforces that FREQ appears first
- allows (but ignores) X-* parts
- improved validation on various values
- error messages are more specific
- enforces that only one of UNTIL and COUNT may be present [disabled]
- allows lower-case property and enumeration values [disabled]
As part of the transition process, both versions of the parser are
called on every request, and the results are compared. If the results
are different a warning message is logged.
An unnecessary constructor was removed.
This also this moves some EventRecurrence tests out of CalendarProvider,
into coretests, and adds a simple parse test with the examples from
the RFC.
Bug 4575374
Change-Id: If737ed1272fda65c93363d87b2da12b85e644f5b
DHCP renewal can finish pretty quick most times and holding
a timed wakelock wastes battery.
Hold a non-reference counted lock and release it immediate
while having the safety of a 40s release
Bug: 4575773
Change-Id: I49d2e8c27b2723690aef26417c4b9a145843a401
