Also fix the error handling for the property_get.
This is part of preparation for the threadLoop() merge.
Change-Id: I6405190ea18146d1271575e1dfe9f279e8f36b17
re-add support for pixelformats L_8, LA_88 and RGB_332 in libui
for backward compatibility.
This may or may not fix 6058926
Bug: 6049685
Change-Id: Ic1b8b4cc994522f7fe664da64c0ef76b98bc6d53
My previous change to speed up the time the IME is dismissed was
fundamentally flawed. That change basically switched the order
the application called the input method manager service from doing
startInput() and then windowGainedFocus(), to first windowGainedFocus()
and then startInput().
The problem is that the service relies on startInput() being done
first, since this is the mechanism to set up the new input focus,
and windowGainedFocus() is just updating the IME visibility state
after that is done. However, by doing the startInput() first, that
means in the case where we are going to hide the IME we must first
wait for the IME to re-initialize editing on whatever input has
focus in the new window.
To address this, the change here tries to find a half-way point
between the two. We now do startInput() after windowGainedFocus()
only when this will result in the window being hidden.
It is not as easy as that, though, because these are calls on to
the system service from the application. So being able to do that
meant a fair amount of re-arranging of this part of the protocol
with the service. Now windowGainedFocus() is called with all of
the information also needed for startInput(), and takes care of
performing both operations. The client-side code is correspondingly
rearranged so that the guts of it where startInput() is called can
instead call the windowGainedFocus() entry if appropriate.
So... in theory this is safer than the previous change, since it
should not be impacting the behavior as much. In practice, however,
we are touching and re-arranging a lot more code, and "should" is
not a promise.
Change-Id: Icb58bef75ef4bf9979f3e2ba88cea20db2e2c3fb
The bit-field TrackBase::mFlags was supposed to have track-specific
flags in the upper 16 bits, and system flags in the lower 16 bits.
The upper 16 bits of mFlags were initialized in the TrackBase
constructor from the flags parameter of IAudioFlinger::createTrack()
and IAudioFlinger::openRecord(), and the lower 16 bits were cleared.
However, the upper 16 bits of mFlags were never acccessed again.
So really there are no track-specific flags. I left the flags
in the parameter list of createTrack() and openRecord() but made a
note that these should be removed eventually as they are dead.
This leaves only the one system flag "step server failed". I replaced
the bit-field mFlags by bool mStepServerFailed, which is simpler and
slightly faster.
Change-Id: I6650f5487be72791b4a67d73adcd10ffa04e2aa5
This change introduces a few new bits of data on
Notification that will help the Notification Manager and
System UI route and display them more intelligently:
-> priority: an integer in a predefined range that
indicates the app's best guess as to the relative
importance (to the user, right now) of that information
-> kind: a tag (really, set of tags) indicating the general
type of notification (realtime, asynchronous, etc)
-> extras: a Bundle of additional key/value pairs
associated with this notification (currently @hidden)
The notification manager takes these data into account when
assigning to each notification a score which is passed with
the notification on to the system UI, where it can be used to
affect presentation. For example:
- Spammy apps (identified explicitly by the user or by
some other means) will have their notifications scored
very negatively by the notification manager, allowing
the UI to suppress them
- Notifications of higher score might be shown larger
or in a different way
- Very important notifications (indicated by a very high
score) might interrupt the user during an otherwise
important task (videochat, game, etc)
Implementation note: This replaces/extends the old internal
notion of "priority", which was mostly used to organize
ongoings and system notifications at the top of the panel.
Change-Id: Ie063dc75f198a68e2b5734a3aa0cacb5aba1ac39
1. Before there were two caches one in the app process that
kept track only the ids of infos that were given to a
querying client and one in the querying client that
holds the infos. This design requires precise sync
between the caches. Doing that is somehow complicated
since the app has cache for each window and it has
to intercept all accessibility events from that window
to manage the cache. Each app has to have a cache for
each querying client. This approach would guarantee that
no infos are fetched twice but due to its stateful nature
and the two caches is tricky to implement and adds
unnecessary complexity. Now there is only one cache in
the client and the apps are stateless. The client is
passing flags to the app that are a clue what nodes to
prefetch. This approach may occasionally fetch a node
twice but it is considerably simpler and stateless
from the app perspective - there is only one cache.
Fetching a node more than once does not cause much
overhead compared to the IPC.
Change-Id: Ia02f6fe4f82cff9a9c2e21f4a36747de0f414c6f
Avoid using wp<>::unsafe_get() except in a log, and other specific cases
when it's known to be safe.
Use more specific subclass types for parameters to avoid down-casts.
When a constructor or method parameter is "this" of an object that is
currently being constructed, it's better to use a raw pointer rather
than either sp<> or wp<>.
Using the raw pointer is safe, provided either:
- it is "this" of an object being constructed (which has sp<> refcount of 0),
- or the caller already holds an sp<>
The raw pointer is simpler and faster, and it avoids the problem of the
sp<> reference count being incremented and then decremented to zero on
scope exit, which would cause the object's destructor to run while the
object is still being constructed.
Also removed some dead code per a review comment.
Change-Id: I7375f64da3aec11b928c33cb01faff186252ef5e
warning: pointer of type 'void *' used in arithmetic
warning: enumeral and non-enumeral type in conditional expression
Change-Id: I7b8d626a636145ef648e3b5d0e77068216dd012e
1. UiTestAutomationBridge was accessing the root node in the
active window by tracking the accessibility event stream
and keeping the last active window changing event. Now
the bridge is stateless and the root node is fetched by
passing special window and view id with the request to
the system.
2. AccessibilityNodeInfos that are cached were not finished,
i.e. not sealed, causing exception when trying to access
their children or rpedecessors.
3. AccessibilityManagerService was not properly restoring its
state after the UI automation bridge disconnects from it.
I particular the devices was still in explore by touch mode
event if no services are enabled and the sutomation bridge
is disconnected.
4. ViewRootImpl for the focused window now fires accessibility
events when accessibility is enabled to allow accessibility
services to determine the current user location.
5. Several missing null checks in ViewRootImpl are fixed since
there were scenraios in which a NPE can occur.
6. Update the internal window content querying tests.
7. ViewRootImpl was firing one extra focus event.
bug:6009813
bug:6026952
Change-Id: Ib2e058d64538ecc268f9ef7a8f36ead047868a05
Two things: (1) make sure the boot message is always positioned within
the entire unrestricted display, and (2) allow the dim background to go
on top of the nav bar when being used for the boot message (this latter
is really a hack that should be more generally fixed in the future).
Change-Id: I7261b044eb802a39cadff931b50a679ff18781d6
- Move mPolicy.startAnimationLw and mPolicy.finishAnimationLw into same method as mPolicy.animatingWindowLw.
- Fix first parameter of performLayoutLockedInner(initial, ...) to pass true on initial pass.
Change-Id: If1b47bb8a7e03cf427769c657e371abc0910b3e3
Use PowerHAL to set system awake/suspend state.
Change-Id: If58a6f548564ea141b68f304455997d9ff04eace
Signed-off-by: Todd Poynor <toddpoynor@google.com>
There are occasional races during application shut down where the
input dispatcher will close an input channel before the application
has finished its last event. So just ignore EPIPE.
Also tweak the logging for failed input event injection to make
it clearer which pid was trying to perform the injection.
Bug: 6013004
Change-Id: I7bbb01441d41762b03eafd4d39dcf0323e1cadf3
Change clock sync control to velicity form PI loop. Tuned for office LAN and
WiFi conditions, will probably perform better in clean environments.
Improve packet filtering to prevent clock sync on bad rtt.
Changed diag interface to take rtt times, P, I, D are no longer supported.
Change-Id: Iad2b26eb44cd222ec5f219b49669e2d6baec9d1c
Use DefaultKeyedVector::valueFor to avoid extra test
Make local variables as local as possible
No double parentheses
No typedef for single use
No parentheses around indirect function call
No AudioFlinger:: prefix when not needed
Remove unnecessary casts
Remove block with only one line
Saves 128 bytes
Change-Id: I3a87430eeb01b81e7b81a1c38f6fdd3274ec48f3
Add a bandaid to prevent a segfault which can occur while handling
timed audio buffers. There is a deeper problem which should
eventually be addressed, but for now this fix should prevent any
crashing.
The deeper problem is as follows.
When the AudioFlinger mixer gets data to mix from an AudioTrack, it
ends up getting a structure filled out which points into an IMemory
region owned by the AudioTrack. Unfortunately, this structure is not
holding a refcount on the IMemory which it points into. If the
IMemory refcount hits 0 and the chunk of RAM is retuned to the binder
heap it came from, there can still be a Buffer object being held by
the AudioFlinger mixer which points into the region of memory which
was retuned to the binfer heap. If AF reads from this buffer, it
could read corrupt data (if the region of memory gets handed back out
to a writer), or it could segfault (if the heap has been freed and the
pages unmapped). Similar problems could happen if AF attempts to
write to the buffer, heap corruption in one case, segfaulting in the
other.
In the past, this has not been an issue for AF, because tracks
allocate a single IMemory (which serves as a ring buffer) and the
IMemory lives for as long as the track lives. As an artifact of the
way the code came out, the mixer cannot be holding a Buffer structure
pointing into the IMemory which used to be owned by a track if the
track no longer exists. Tracks cannot come into or out of existence
during a mix operation, which is the only thing which makes this safe.
TimedTracks work differently, however. Timed tracks each allocate a
small binder heap, and then hand out IMemory instances broken out of
this heap. The heap lives as long as the track, so the worst which
could happen here is that a TimedTrack's IMemory gets returned to the
heap while there is still a buffer structure in flight pointing into
the memory region, then the region gets handed out again and
overwritten by new data causing the mixer to mix the wrong audio. The
timing to cause this to happen is very difficult to encounter, and you
to generate the timing conditions required, you need to be in a pretty
bad failure state where audio is already breaking up and skipping, so
its unlikely that anyone would notice (which is why I'm band-aiding
the segfault and letting the deeper issue slide for now).
In general, however, it might be a good idea to revisit this buffering
design. On principal, if someone is going to hold pointers into a
refcounted object, they should be holding a ref on the object at the
same time. Failure to do this will usually lead to a situation where
there are corruption or segfault issues, or to a system where the
refcounted object's lifetime must be implicitly managed very carefully
in ways which are usually non-obvious and are easy to break by new
engineers on a project.
Change-Id: Ib391075395ed0ef46a03c37aa38a82d09e88abeb
"EffectDesc *effect = new EffectDesc(*effects[i]);" was relying on the
default copy constructor for EffectDesc, but the default copy constructor
does a member-by-member copy. This works OK for mUuid, but a member
copy of mName and mParams shares pointers. This could result in heap
corruption later on due to a double free. Changed to add an explicit
copy constructor that does a deep copy of both mName and mParams.
A malloc() and strdup() were being freed by delete, but the correct
matching API for these is free(). Fortunately our current memory runtime
implementation ignores the difference. Changed to use free().
EffectDesc and InputSourceDesc member fields were being torn down by
the code that does delete. Changed to do the tear-down in ~EffectDesc()
and ~InputSourceDesc().
Added constructor EffectDesc() with name and UUID parameters, rather
than having caller fill in the object after construction.
Made ~EffectDesc() and ~InputSourceDesc() non-virtual to save memory,
since they have no subclasses.
Change-Id: Ibb5cc2e6760d72e0c4cf537068ac4432c717bafd
