fix(#515, #325): review round 3 - bare-day EXDATE/RDATE on timed series resolved when read, unaccepted occurrences of declined series not copied

This commit is contained in:
makiolaj committed 2026-11-14 03:22:45 +01:00
1 parent 05887d2ed0
commit 1e83000696
7 files changed
+63 -20

No files matched your search

+5 -2
View File
@@ -522,7 +522,9 @@ as before.
alone. The provider stores `RDATE` in the same `<zone>;<stamps>` form as
`EXDATE`, so `exportRecurrenceDates` resolves it the same way to UTC stamps (or
bare days for an all-day series), line by line where several `RDATE`s were
stored; a `PERIOD` entry keeps its start. `IcsWriter`
stored; a `PERIOD` entry keeps its start, and a bare day on a timed series
becomes the occurrence at the series' time of day, so the copy (which never
goes through the parser) and a file get the same stamps. `IcsWriter`
writes it as `RDATE:` / `RDATE;VALUE=DATE:`, anchors such a series to its zone
like an RRULE series, and writes its `EXDATE` too; `IcsParser` reads `RDATE`
through the same normalisation as `EXDATE`, so a foreign `TZID` or floating
@@ -535,7 +537,8 @@ what landed before. It skips the `.ics` text: the writer pins a timed one-off to
UTC and carries no colour or visibility; the direct mapping keeps them.
Attendees and organizer are never read, and cancelled events and declined
invitations are left out (an insert can't be cancelled; it would arrive live, and
a declined occurrence counts as deleted). A re-run is idempotent per
a declined occurrence counts as deleted, as does one of a declined series the
user didn't accept on its own). A re-run is idempotent per
snapshot: a series already in the target keeps its overrides, so an occurrence
changed since keeps its earlier copy rather than gaining a second one.