Time Change, Date and Treatment Discrepancy

[marionbarker]

Describe the bug I am noting discrepancies in date vs some treatments (and thus plots) before and during the recent time change. I believe it is basal and events that don't align as expected.

Comment I'm happy to close this issue and add to an existing issue if that is preferred - just let me know which issue number. I believe this is related to a long-standing problem when the computer time is not aligned with the profile time. I believe this was partially fixed with release 15.0.

Configuration

• Nightscout 15.0.2
• Uploader: Loop, dev branch (commit 6b7bba6) on an iPhone 11 pro, iOS 17.1.1

Details and Graphics The graphics are way zoomed out to show multiple days, so are poor resolution - just look at the overview. I describe the steps to generate each graph and then follow up with composite graphics to emphasize the discrepancy.

For all reports, I started fresh (close report, open report and configure) so that prior fetches from the database are cleared.

For the US, the time change was at 2 am on Nov 5, 2023 (11/5/2023 using US notation.)

First test, all days are before time change:

• Open report, select day to day from 11/1 to 11/4
  • first time hit of submit, plots labeled 10/31 to 11/4 are displayed
    • only first day (10/31) shows basal plot and "default" at midnight
    • subsequent days (11/1 through 11/4) show no basal plot and "default" at 23:00
  • second time hit submit, only the requested days are plotted (11/1 through 11/4), still with no basal plot and "default" at 23:00
  • extract just the plots for Nov 2, 3 and 4 and add to composite below (right side, top)

Second test, select days that span the time change:

• Open report, select day to day from 11/3 to 11/6
  • get same result every time you hit submit, plots labeled 11/2 to 11/6 are displayed
  • all show basal in plots, all show "default" at midnight
  • extract the full display and add to composite below (left side)

Third test, select days after the time change:

• Open report, select day to day from 11/6 to 11/8 (all after time change)
  • get same result every time you hit submit, plots labeled 11/6 to 11/8 are displayed
  • all show basal in plots, all show "default" at midnight
  • CGM Sensor Start of 11/7 shows up on the right date
  • extract just the plots for Nov 6 and 7 and add to composite below (right side, bottom)

First composite graphic - plots on left and right are for same date from different selections for date range for Day to Day plots.

• Please note the upper right plot does not show basal and has Default showing at 23:00

Next comment covers using an Event Filter.

[marionbarker]

Continuing the bug report in a new comment. (Edited to fix date error. CGM sensor started 10/28 not 10/29 as originally stated. Easy to get confused when reported dates do not align.)

Fourth test, search by Event Type vs Select known day:

These days have a CGM Sensor Start in the Treatments:

• 10/28 : start at 2:42 PM with Daylight Savings Time
• 11/07 : start at 1:56 PM with Standard Time

Search using Event Type: CGM Sensor Start:

Start a new report (clears data already loaded)

• Day to Day: Select range 10/28 to 11/09/2023 and Select Event Type: CGM Sensor start
  • plots for 10/28/2023 and 11/07/2023 are shown (sensor start was on 10/29/2023)
    • Sensor start was on 10/28, but data shown was from previous day, with sensor start being displayed on 10/29 plot
  • include resulting plot as left side of composite in graphic below
  • (note - when doing this for treatments, get the same result, 10/28 and 11/07 are returned)

Start a new report (clears data already loaded)

I had the date wrong when I did these tests - what shows up on 10/29 for sensor start actually happened 10/28

• Select date range of 10/29/2023 to 10/29/2023 without adding event type
• First submit shows plots for 10/28 and 10/29
• only first day (10/28) shows basal plot and "default" at midnight
• 10/29 shows no basal plot and "default" at 23:00
• Second submit shows same plot as first submit but only for 10/29
• include first submit plot as right side of composite in graphic below

Composite graphic:

• (left side) the CGM event is correctly selected after the time change whereas the date before the event is selected before the time change
• (right side) the plot for 10/29 has the event (which was actually on 10/28) and does not display the basal or midnight appropriately
  • Note the caption says event happened on 10/29 but actually happened 10/28 - label for plot that contains the CGM sensor start is 10/29 (off by 1)

[stucka]

Possibly related: I have an instance of 15.0.2 at nightscoutpro dot com, which I believe may be on UK time. So it's the night of Monday, January 29, here, and early Tuesday, January 30, there.

Reports: Daily Stats: Show reports entries for Tuesday, January 29, and Monday, January 28.

I can work around the timezone difference but there's a basic discrepancy those dates in the different formats.

[rpm5099]

I am noticing that when I enter treatments in nightscout using the + that after 8pm EST is always defaults to the next day, and if you don't manually change it then it will enter it 24 hours in the future. I think that issue is related, probably the same cause, but I have not had a chance to track it down yet.