All-day events always use UTC, so when the multiday detection code was
converting them to the local timezone, they span an additional day if
the local timezone is not the same as UTC. This was leading to all-day
events always being considered multiday events.
Fix this by ignoring the timezone for all-day events, like all other
comparisons do. Also don't consider all-day events spanning 1 day to be
multiday since as per RFC 5545 the end date is non-inclusive and if
omitted to be start + 1 day.
Zero length ranges, such as for events without a DTEND could be
considered to overlap a range if they were at exactly the end of the
range. The end of a range however is the first date that is supposed to
not overlap, so a zero-length range on exactly this point should not be
considered overlapping.
This fixes cases where all-day events from the next month could show up
on the same day of the current month if they did not have a DTEND.
This also adds tests for all possible cases of such zero-length events
overlapping.
Fixes https://gitlab.gnome.org/GNOME/gnome-calendar/-/issues/701
On systems without anything providing the org.freedesktop.timedate1
interface, the timezone returned by the monitor would always be NULL,
because it was never set. This can lead to crashes in places that expect
the monitor to always return a valid timezone. So in that case use the
local timezone from when the monitor got initialized.
Fixes https://gitlab.gnome.org/GNOME/gnome-calendar/-/issues/703
Sebastian Keller