All background work is going through AsyncTasks, which uses a shared
thread pool. Even with the new ContentProviderClient logic to detect
ANRs, the UI can still appear to be unresponsive for 20 seconds, even
if the user attempted to switch to a different backend. In the worst
case, a backlog of thumbnail requests would end up wedging Loaders
for a long time, since they all share the same THREAD_POOL_EXECUTOR.
This change isolates calls to each provider onto their own thread,
which they're free to wedge and recover from over time.
It also means we no longer need a dedicated thread pool for recents
loading, and can use a simpler Semaphore instead.
Disables thumbnails in recents on svelte devices.
Bug: 10993301, 11014856
Change-Id: I7f8a5bbb5f64437e006cb2c48b7e854136d5c38c
The main problem here was a mistake when turning a single process
structure to a multi-package-process structure with a common
process. When we cloned the original process state, if there were
any services already created for the process for that package, they
would be left with their process pointer still referencing the
original now common process instead of the package-specific process,
allowing the active counts to get bad. Now we switch any of those
processes over to the new package-specific process.
There was also another smaller issue with how ServiceRecord is
associated with a ServiceState -- we could be waiting for an
old ServiceRecord to be destroyed while at the same time creating
a new ServiceRecord for that same service class. These would share
the same ServiceState, so when the old record finally finished
destroying itself it would trample over whatever the new service
is doing.
This is fixed by changing the model to instead of using an "active"
reference count, we have an object identifying the current owner
of the ServiceState. Then when the old ServiceRecord is cleaning
up, we know if it is still the owner at that point.
Also some other small things along the way -- new Log.wtfStack()
method that is convenient, new suite of Slog.wtf methods, fixed
some services to use Slog.wtf when catching exceptions being
returned to the caller so that we actually know about them.
Change-Id: I75674ce38050b6423fd3c6f43d1be172b470741f
Glow absorb animation was very slow with high velocity flings. Clamped the velocity
and adjusted constants for start values and duration of absorb animation.
Change-Id: Id2a87814498e05f5f8ceb942dd1d765874461beb
All ContentProvider calls are currently blocking, making it hard for
an app to recover when a remote provider is wedged. This change adds
hidden support to ContentProviderClient to timeout remote calls,
treating them as ANRs. This behavior is disabled by default.
Update DocumentsUI to use a 20 second timeout whenever interacting
with a storage provider.
Bug: 10993301, 10819461, 10852518
Change-Id: I10fa3c425c6a7225fff9cb7a0a07659028230cd3
When an inset computer is registered with the view tree observer,
we report content insets to the window manager. When an inset computer
is subsequently unregistered, we must take care to clear these insets.
This patch remembers whether the previously computed insets were
non-empty so that it can force insets to be reset when needed.
Bug: 10840662
Change-Id: I4cce5ba64cc5234b98363b025ac4bb42e64349f1
This ignores any previous setting and instead uses
a value set at build time. This does not preclude
us from using some other signal to determine country
for wifi channel limits.
bug:10513734
Change-Id: Ib82c07285af70fbd82eb0466b7391979ebc8be10
While the logic was correct the array map that holds the list of pritners
does not keep the position of the items constant. Switched to linked hash
map which gives this guarantee.
bug:10955751
Change-Id: Idbbe14d753e6a1ad1002f2289b10cb62d7f9f040
When loading historical records for previously used printers we
now ignore the ones whose target print service is not installed.
bug:10955652
Change-Id: Ib295e7d88ed3c308ef6d8a11bdc1792ebbb6d526
1. We want the files for a print job to be removed as early as possible
typically because the print job was cancelled, completed, the app
or the spooler crashed during print job construction. We were keeping
around in the spooler and hence to disc infos for jobs that are in
final state since the app that created them may hold a reference to
a local print job objec whose info it can access to get the latest
print job state potentially after the job reached final state. The
issue was that we were persisting to disc created print jobs which
were during construction which requires careful handling for the
various cases above. This is tricky and error prone.
We used to tell the spooler to forget the print jobs infos when the
app that created them died. The implementation to forget a print
job was not careful and was nuking currently running print jobs in
addition to the ones in a terminal state. Further, if the app dies
before a print job is completed we were left with a stale print
job in the spooler since we missed the signal to forget it (assuming
we forget only inactive jobs). These issues suggest that the approach
is problematic.
Now we have a cache of print job infos for the jobs an app created.
This cache is updated when the state of a print jobs changes using
the new print job state observation code. When the app dies we
remove the cached jobs for that app. Now if the app calls to get
the print jobs it gets the cached ones, i.e. the print jobs it
created during its lifetime, plus the print jobs that are still
active fetched from the spooler. Note that transient state cannot
be kept in the spooler since we unbind from it if there is no
work and it may get killed.
2. Improved the print sub-system logging code to show the cached
print job infos for apps and also dump the print job PDF file
names.
bug:10958357
Change-Id: I6f7c1968b6b7ba5be182a10df044ff7ea1fc3a61
Yet another iteration from UX on how roots should be ordered. Since
we no longer categorize by type, remove from public API. Updated
asset drop with new dividers.
Update public API docs to be explicit about required columns. Hide
flags and columns that aren't required for third-party apps.
Move remainder of potentially blocking work to AsyncTasks, including
creating directories, picked root resolution, and creation of new
documents once picked.
Improve performance of layouts by removing baseline alignment and
reduce hierarchy depth. Set alpha on ImageViews directly to avoid
offscreen rendering hit.
Limit returned recents to 45 days. Show load in recents when still
waiting for backends. Show empty message when no recents stacks to
create from. Use unique key when saving recent stacks.
Bug: 10941423, 10819454, 10964412, 10960718
Change-Id: I08cf589dcda7e203acf67928f4d30322ae36ee94
Since binder call permissions are not transitive by design,
the proper way to fix this is to have the call talk directly
to keyguard from the navigation bar.
Fixes bug 9409008
Change-Id: Ibd90a79bb638c969b514455a2ad93c6ff668222d
Add the concept of an anchor looper.
Allow show/hide/selectTrack/selectDefaultTrack to be called from thread
other than anchor looper.
Change-Id: Iaf2f3d3c46d1b00a0a24156745ec16b301c80c04
Signed-off-by: Lajos Molnar <lajos@google.com>
Bug: 10919206
When in the SyntheticKeyHandler, a missing constructor paramater caused the
wrong contstructor to be called. This would cause the device id, scan code and
flags to be incorrect.
Bug: 10966229
Change-Id: Ib84ef88d28c0da517b376c5493a851cf8fe0a9f7
This is a fix for bug 10863979.
The onPreDraw() method in TextView was canceling the draw every time a
the horizontal scroll changed. These dropped frames were visible as
jank. The fix is to never cancel the draw. It seems likely that the
previous rationale for canceling draw is no longer valid, due to other
fixes, because the scroll state will have been fully updated.
Change-Id: I8c2da5450723f993055a49b1cb57320c08f2a11e
- Calling build() on a Style now goes through the same
codepath as calling build() on the Builder.
- Documented new constants and unhidden classes.
- Fixed crash in Action.clone().
Bug: 10112103
Bug: 10461196
Change-Id: I08cd94790b538a361ccf8ff3682f6a86a7812b95
