- 20 Jan, 2016 1 commit
-
-
Justin Klaassen authored
Bug: 26484535 Change-Id: I32e8f1c5f8620109affc25e3bb4a826856759eec
-
- 19 Jan, 2016 1 commit
-
-
James Lemieux authored
Bug: 26591272 Prior to dazzle, timers could inhabit a state of "DELETED" (value = 5). In dazzle and beyond there is no reason to preserve such a state, but it is possible for timers to exist in SharedPreferences in such a state, so it must be handled gracefully. Change-Id: I1754ef898dfff84c0a3e807c4bd0dd1473f87aad
-
- 14 Jan, 2016 1 commit
-
-
James Lemieux authored
Bug: 25846551 The original form of the code that fires an alarm is: AlarmManager -> BroadcastReceiver -> Service. The new form of the code that fires an alarm is: AlarmManager -> Service. Evidence exists that the system lowmemorykiller may elect to kill the clock app after BroadcastReceiver.onReceive(...) completes but before Service.onStartCommand(...) begins. When this occurs, the results are disastrous as the clock fails to fire at the appropriate time. To remove this possibility, all alarm state changes are delivered to the Service. The methods that manipulate database state within BroadcastReceiver have been made public and static and are called directly from the Service to perform the same work as before. If the alarm state transition is to the FIRING state, the AlarmService also performs the work of posting the firing notification in the foreground. All of this occurs during the handling of a single Intent on the same thread within the service which should no longer provide lowmemorykiller with any opportunities to prevent the firing of an alarm by killing the clock app. Change-Id: I3629a5b725a758f680f41611939a5bbeec23238a
-
- 12 Jan, 2016 1 commit
-
-
James Lemieux authored
Bug: 26471891 This makes the clock app unlikely to be killed in memory pressure situations while expired timers are ringing. Change-Id: I89f141a835e3de67a58671d6c5b381de52be5c18
-
- 18 Dec, 2015 1 commit
-
-
Justin Klaassen authored
Bug: 25084547 Change-Id: I62df341b463d976264bf5fa84dbc5c22b6d1cd4c
-
- 15 Dec, 2015 2 commits
-
-
Dylan Phan authored
-
Dylan Phan authored
Bug: 26190910 Change-Id: Ie0c844708d51b04be504c8dd68e358e017f2e28f
-
- 14 Dec, 2015 1 commit
-
-
Geoff Mendal authored
Change-Id: I9857577797267755f5b634732d76e597f643042b Auto-generated-cl: translation import
-
- 10 Dec, 2015 1 commit
-
-
James Lemieux authored
-
- 09 Dec, 2015 2 commits
-
-
James Lemieux authored
Bug: 26058053 Evidence shows that AlarmService is being destroyed a couple 100 milliseconds after starting. This can be catastrophic as it typically means that the ringtone playback is ended before it even sounds. This CL balances the binding/unbinding calls in AlarmActivity to AlarmService to encourage the UI to stay in the foreground and keep the service alive and the ringtone playing. This CL also removes unnecessary cancel()ations of notifications. Change-Id: I59b35dc8404dbd1f8d74285aa34a7cd81ec5545b
-
Geoff Mendal authored
Change-Id: I99b0010be7f1b3fd8bae24147e80acbdd5f0c71f Auto-generated-cl: translation import
-
- 08 Dec, 2015 1 commit
-
-
James Lemieux authored
Bug: 22371346 This is a use case that can only occur via system intents and not via the user interface. Since the utility of pausing and/or resuming an expired timer is dubious at best, the effect is now changed to reset the expired timer. Change-Id: I28c7500d1393dcfade1303b31de750c525dd0df3
-
- 07 Dec, 2015 1 commit
-
-
Geoff Mendal authored
Change-Id: I142ba61896ebddb20aa7dd76c1c555fb89fee99f Auto-generated-cl: translation import
-
- 02 Dec, 2015 1 commit
-
-
James Lemieux authored
Bug: 25895834 ag/811037 left a bug in place that this change addresses. Specifically, if a one-time alarm is missed, an entry is left in the alarm_instances table with a state of MISSED. The query's join clause would not return a row for the alarm in this case. A subselect is now used to return the alarm_instances row with the smallest state value. This means: - SILENT instances are preferred over MISSED instances (the only case where 2 instances exist for the same alarm). - MISSED instances are returned if they are the only instance for an alarm (the above case) Change-Id: I31de6d8b2d2bf7719f6b027a2401b1da1a09f247
-
- 01 Dec, 2015 4 commits
-
-
Justin Klaassen authored
-
Dylan Phan authored
-
Justin Klaassen authored
Bug: 24341005 Change-Id: I558c6d15316c9391ae57b96b7324ebdf1eddc29e (cherry picked from commit 09e29779)
-
Dylan Phan authored
Bug: 25695607 Change-Id: I41c20a905566f36827b433631baee97362ad7d02
-
- 30 Nov, 2015 1 commit
-
-
Geoff Mendal authored
Change-Id: I93002905a560686d0bbdc6535c4fe3506da45526 Auto-generated-cl: translation import
-
- 26 Nov, 2015 1 commit
-
-
Dylan Phan authored
-
- 25 Nov, 2015 5 commits
-
-
James Lemieux authored
-
James Lemieux authored
TimerIntentsTest actually caught this bug. The intent to show the timer tab requested that a freshly deleted timer be displayed. This can produce exceptions if it was the last timer. When a timer is deleted, the timer tab can be shown but no specific timer should be requested for display. Change-Id: I7024fbe5a154d5ecb8d563c80c09044d3917ab50
-
Geoff Mendal authored
Change-Id: I573d63871f3c2ad4949131c9419a32ff0f24b8b9 Auto-generated-cl: translation import
-
James Lemieux authored
-
James Lemieux authored
In order to sidestep b/25872174 it was agreed that code attempting to show the "Night mode" menu item on just the clock tab would be removed. Change-Id: I842b093c7d206dc8e42a6a85bbd487dd9a235942
-
- 24 Nov, 2015 2 commits
-
-
Dylan Phan authored
Bug: 24496589 Change-Id: Ic79b44e47824e3a1a494f28b87d2dde9a0a0d9f9
-
Dylan Phan authored
Bug: 24496589 Change-Id: I3bc06064fd13dfdefc91db04cfe678ee7e840645
-
- 23 Nov, 2015 4 commits
-
-
James Lemieux authored
-
James Lemieux authored
Bug: 25843144 Attempting to view a specific timer trumps the timer setup screen in all cases. onResume() now enforces that policy. The timer setup screen is still displayed in all other cases, including: - no timers exist - the intent indicates a new timer should be created - the last visible timer screen was the timer setup screen Change-Id: I28d147cadeb29055fa0c68a44da6648f6ad33238
-
James Lemieux authored
-
Geoff Mendal authored
Change-Id: Ic357a1d328f780b68a8c9ff48d49086f830c6d9e Auto-generated-cl: translation import
-
- 21 Nov, 2015 1 commit
-
-
James Lemieux authored
Bug: 25302368 The bug was already fixed as part of b/19651361 but I wanted to add the reported use case to the corpus for better regression testing. Change-Id: I51f80cfbbae01681b018d249de56d0d39b4122d2
-
- 20 Nov, 2015 2 commits
-
-
James Lemieux authored
-
James Lemieux authored
Bug: 25780110 Change-Id: I454ee709d296c7a2860281b2e981476608df1e57
-
- 19 Nov, 2015 3 commits
-
-
James Lemieux authored
Bug: 25779786 The issue here is that TimerSetupView actually customizes the fab. These customizations include setting it invisible. It was a poor design to share the fab with this view. So, the fix is that the view now holds its own fab-like button that it is free to manipulate without fear of altering the app-wide fab state unintentionally. Change-Id: Ie83b03c98db4f77e0be912b122b6f74c2c7e6c8a
-
James Lemieux authored
-
James Lemieux authored
- add precision to some wording for translators - update widget provider XML layout Change-Id: I78d7d29b2ed2ea4768c0e36d125b38df971c1484
-
- 18 Nov, 2015 2 commits
-
-
Geoff Mendal authored
Change-Id: Ife7cf799c6d9ca5cf15deff3337526295fdf19b5 Auto-generated-cl: translation import
-
Justin Klaassen authored
-
- 17 Nov, 2015 1 commit
-
-
James Lemieux authored
Bug: 24977472 By default, the menu is inflated after the first draw. A proactive call to Toolbar.inflate(...) during creation defeats this default behavior. A second problem is that the menu that is inflated by default for DeskClock contains the search menu item which is briefly shown. By defaulting that menu item to be invisible and only making it visible when its corresponding controller exists, the dancing/flashing effect is avoided. Change-Id: I933ab9a59b471abe26e1cf07385e4a4c128d1808
-