Tasks that start one day and finish the next do strange things in reports.

[mdtrooper]

I have this example:

The left panel in the terminal is a watson log and the right panel is several watson report for each day:

There was a task in "Saturday 14 February 2026" "fhe:lp_mantenimiento_12_parejas_bug" that it takes 4h49m of the night (yes I needed to do a crunch time in the work) and the next day I worked again in the same task "Sunday 15 February 2026" "fhe:lp_mantenimiento_12_parejas_bug".

Well the sum of the times of all task of these 2 days is correct .

But the watson report -G -c -f 2026-02-14 -t 2026-02-14 shows the "fhe:lp_mantenimiento_12_parejas_bug" task only with the hours of this day. And the watson report -G -c -d (on sunday, I am writing the issue the next monday) shows the sum the rest tasks from fridays night (well saturday 00:00 to 04:00).

I think that watson report needs a parameter to don't cut the tasks that starts on a day and finish the next day. Or watson log need a parameter for grouping the equal tasks in the same day.

[hiSandog]

The cross-midnight case needs an explicit reporting semantic. Users may expect either attribution to the start day, attribution to the end day, or splitting the interval across calendar days; each is defensible but reports should be consistent. A regression test with a task spanning midnight, and ideally a DST boundary too, would make the chosen behavior visible and stable.