Alarm app sometimes reports thousands of days until next alarm

[markmatney]

• I searched for similar bug reports and found none was relevant.

What happened?

On a Sunday, after using a companion app to synchronize the watch time, the alarm app reported thousands of days until the next 08:00 MON-FRI alarm.

What should happen instead?

The alarm app should have reported less than a day until the next alarm.

Reproduction steps

1. Circa 15:00 local time: boot up watch after recharging from dead battery
2. Upgrade to v1.6.0 from v1.2.0 using Siglo
3. Open alarm app, set alarm to 08:00 MON-FRI, toggle alarm ON:
   
4. Close alarm app
5. Sync watch time with laptop computer using Siglo
6. Circa 15:10 local time: open alarm app, press ℹ️ button in alarm app, note very long time (~ 84 years) before next alarm:
   
7. Circa 15:12 local time, toggle alarm OFF, toggle alarm ON, press ℹ️ button, note expected time before next alarm:
   

More details?

No response

Version

v1.6.0

Companion app

Siglo v0.8.12

[mruss77]

Thanks for reporting! This is an issue with how the app currently sets the alarm in conjunction with time changes on the watch. I'm working on an improved alarm app that will fix this and add new features like multiple alarms and saving the alarms through reboots. Currently the alarm app does not behave correctly when the time is changed on the watch.

As for why it happens - at the time you set the alarm, the app calculates and stores the date/time when it should go off and schedules a timer to wake up the app at that date/time and trigger the alarm.

When you press the "i" button, it calculates the time difference between the current date/time and the alarm date/time.

In your scenario, the date and time change (probably by quite a lot) between when you set the alarm and when you check the time-to-alarm. The time difference is now negative but is translated to an unsigned integer format, which results in this very large time difference you're seeing. I don't know when, if ever, the alarm would actually go off since the timer that was set is for a time that is now months or years in the past.

Until this is fixed, it can be worked around by turning the alarm off and back on when the time is changed on the watch.

[Riksu9000]

I believe this could be fixed by recalculating the timepoint if the alarm is set here.

InfiniTime/src/systemtask/SystemTask.cpp

Line 273 in 71ce13d

case Messages::OnNewTime:

But also #698 recalculates the timepoint each time, so wouldn't that also fix this?

[mruss77]

@Riksu9000 that seems like a good approach. #698 fixes the time figures displayed to the user but we'd have to also reschedule the alarm.

I'm thinking about making a small patch to Alarm to fix this and persist the alarm setting, because the more complex stuff may take a while.

[brabalan]

This just happened to me: my alarm did not ring this morning at 7, and when I tapped the info button it reported a time very far in the future (as if it was midnight many years in the past). The watch was disconnected from the companion app (GadgetBridge) as my phone was in another room. After connecting the watch to the phone, the time reported was now correct.

I’m using 1.9.0, so I thought this would be fixed. Could something be happening when the watch loses connection to the app? (I tried manually disconnecting but I could not reproduce the issue.)

[TinfoilSubmarine]

I also had this happen on 1.9.0 this weekend with the info button reporting a time very far in future.

[brabalan]

I experimented some more:

• I put the alarm later, so that it would ring after I woke up. The time till the alarm was correct (about 25 minutes in the future), but the alarm did not ring at the expected time. The time till next occurrence of alarm again reported many thousands of days in the future.
• I tried an alarm 2h in the future, it rang correctly and the time to next alarm was also correct.

[JF002]

The original issue seems to have already been fixed, so this issue is either a new one or a regression that appeared after the original fix.
Could you create a new issue (bug) with all the information that could help debugging it?