The old one didn't use the app filter when packages
were added or updated, only in the initial load.
The app filter is now also used for appwidget providers.
Bug: 10895858
Bug: 10898990
Change-Id: Id29bbc3cf7bdb0c360edd92f768b86684bd338ed
Note that this doesn't honor folder nesting (that is, you
can't include an XML file full of shortcuts inside a
<folder> node and expect them all to appear in that folder).
It's really just a way to concatenate/overlay multiple XML
files, making it easier to make a few tweaks/additions for
certain screen sizes or locales.
Bug: 10932771
Change-Id: I1fc29477d0717a8a7bc116f16729a955e6c625d5
-> In this approach, the shown/hidden state should be tracked
by the CustomContent provider and excess calls to onShow and onHide
should be debounced. That is, if onShow is called twice in a row,
without an intermediate onHide, the second should be discarded.
Viceversa for onHide.
Change-Id: I2fa48d15f28767775a9a9cb84c1d920867a12176
We shouldn't animate from here as this only gets called when
transitioning to the custom content screen and we shouldn't
animate exiting overview mode.
Bug: 10849624
Change-Id: I1309f2ed1f89a20a9e7b7a2fb5aec47b01f222ab
Only allow a small number of icons to be backed up in any given pass.
Also refactor common code out of favorite and screen backups.
Bug: 10778984
Change-Id: I54bc769c1d1c1c9087ea4bc58f258bd15c167aea
Whereas standard ShortcutAndWidgetContainers are static during screen
transition, the custom screen may animate during its transition.
Disabling the hardware layer reduces the cost of this animation
significantly.
bug:10810505
Change-Id: I560d209f651951f54c224862a706e8a243dc71af
-Move transparent bars from just GEL to Launcher3
-When wallpaper strip animates, render it under the nav bar
-Disable rotation of wallpaper picker on phones
Bug: 10814785
Bug: 10852650
Bug: 10852554
Change-Id: I9efeccbc4ad1933689266a5dede201ccfd34acf4
-> This was caused by a strange race condition. The page snap time was equal to
the delay to recheck whether to snap (in DragController). This meant that
scrollRight()/Left() would get called, and the scroller would be finished
however, the final computeScrollHelper() hadn't been called, so the mCurrentPage
hadn't yet been incremented.
-> Fixed the underlying race condition.
-> Added suitable gap (150 ms) between the two delays.
Change-Id: If700eb9e14d77a174e4395ca6933119bdb0da768