The CL lets MountService to observe device FD, and request unmount to
vold when the device FD was closed, or remote application providing
appfuse is crashed.
BUG=25756420
Change-Id: I7990694d32affa7f89e3f40badb25098d74d744d
Previously the while loop condition refers to Thread.interrupted()
wrongly. The Thread.interrupted() flag is cleared when the
InterupptedException is thrown.
BUG=26297698
Change-Id: I113d5c88389d54a4a6e29af2cf9515b1693b59dc
Primary goals of this refactor are to reduce DirectoryFragment bloat,
and to simplify the code (especially the binding code) for the different
layouts.
- Decouple DocumentHolder from DirectoryFragment.
- Move it into its own file.
- Move binding code from DirectoryFragment into DocumentHolder.
- Split DocumentHolder implementation into three separate subclasses,
for grid items, list items, and dividers.
BUG=24326989
Change-Id: I217bf4e5b8e1b33173b8b0275591a8c5d8e9161c
When we dock with Pixel C keyboard we normally expect to find it, and so
we perform a low latency scan (which uses highest duty cycle). But if
keyboard is not charged or if tablet mode trigger misfires (because
there was a magnetic object in proximity?) we may not find the keyboard
and will continue scanning indefinitely, impacting WiFi performance and
draining battery. To avoid this scenario let's stop scan if we do not
find "our" device in 30 seconds.
Also remove STATE_WAITING_FOR_STATE_PAIRED as it was unused.
Change-Id: I4bf1a9a163d3165eaa5e8f1cc2d2938465d454af
(cherry picked from commit 6c12c3a6538591b5fa6e11cb4f583aa494e0ce5c)
Signed-off-by: David Riley <davidriley@google.com>
info-activities
The icon is loaded from the discovery session only when it is displayed
to avoid having to store too many icons in memory.
Also the icons are not maintained in the historical printers. Only if
the printers are available nice icons are shown. A historical printer is
updated with the appropriate properties (including icon) once it becomes
available.
Bug: 24135005
Change-Id: Iec389bab514b024634be8fb5fc8928371cba8740
Currently, the bugreport screenshots are taken by dumpstate and passed to
Shell as a path on BUGREPORT_RECEIVED; this change not only delegates the
screenshot taking to Shell, but also allows user to take more
screenshots while the bugreport is being generated.
As a result of this change, the final ACTION_SEND_MULTIPLE intent might
contain multiple screenshot attachments, all of them named
"screenshot-PREFIX-NUMBER.png", where PREFIX is the bugreport
name (either initial date provided by dumpstate or a name entered by the
user) and NUMBER is the sequential number of the screenshot as taken by
the user.
The screenshot is taken using screencap, which not only is simpler than
using Framework APIs, but also faster and less intrusive. The only
drawback is that it might fail if an OEM is not providing screencap; if
that happens in the field, we'll need to add fallback option to do it
using such APIs.
Prior to this change, all work done on BugreportProgressService was
executed in one single thread (through the ServiceHandler class) but the
code was guarded by unnecessary synchronization. Now there is another
thread (ScreenshotHandler) that will be used just for taking the
screenshot (so it doesn't handle the main thread). Despite the addition
of a new thread, the code was simplified to remove most synchronization
locks, excepted for the areas touched by both threads.
Once this change is submitted, the bugreport service will be changed so
it does not ask dumpstate to take a screenshot.
BUG: 26274653
Change-Id: I1df883e3c0ca6e3e3cad2522a6a99585f71abb75
The notifications are now measured at wrap_content such
that the layout still looks good even if the text
was ommited instead of arbitrarily making it bigger.
This also fixes some small rounding issues where the
expanded notification was just a pixel bigger than
the collapsed version.
Bug: 26185377
Change-Id: Ie3339ff75680ab512446605055304576e058f588
Also caching the notification minheight in the algorithm
far less agressively, because it led to bugs.
Change-Id: I57155103bf2700be2e98e686fdd0077ce8e8405b
The notification algorithm had some state which was simply
dropped on rotation and therefore weird things could happen.
This is now fixed.
Change-Id: Ibb3d007b3298f745743bd3a5889bccebebf8105a
Also fixed a bug where the guts didn't have the
right size and in general cleaned up the intrinsic
height. The guts also animate nicely and the position
of their animation starts on the finger.
Change-Id: I62447dd77ace36161740dfdf3d18103524e2eb21
The views were based on a legacy systemUI notification.
We are now creating it in Notification.java in order to
always have the latest visuals and avoiding duplicate work.
Change-Id: I3f544cad069a8eab34d4464404d020e06d7f6626
The notification children didn't respect the given dimensions
when measuring and was simply measuring itself as high as desired.
This lead to a bug where the parent could crash when a layer was
set on it.
We are now measuring the container at most with the height of
the given size and let children draw over our view bounds.
In oder to still allow touch in those regions we also
override the touch rect.
This also simplifies the rest of the touch handling.
Bug: 26159274
Change-Id: I728553a6386455e6632e2511be8a3e7cb447e89b
Change the sort code in the Model to bucket items into two separate
categories (folders and documents) and then sort the two buckets
separately.
Add code to the adapter to insert a visual break in the UI, between
folders and documents.
Change-Id: I759fedcef829aba9ad61554326489a9e62641cc7
The CL updates MtpDocumentsProvider#queryRoots so that it fetches
devices from Database as well as storages when we don't have storages
under the device, or when we have multiple storages under the device.
BUG=26120019
Change-Id: Id2b140f00a1d49fa4da7e17d2564dbbaa1795e1e
