Debian Patches
Status for libical3/3.0.20-3
| Patch | Description | Author | Forwarded | Bugs | Origin | Last update |
|---|---|---|---|---|---|---|
| 32-bit_time_t_upper_bound.patch | icaltime: fix 32-bit time_t upper bound off by two days make_time()'s SIZEOF_TIME_T==4 boundary check rejected any date after 2038-01-17, two days before the actual limit of a signed 32-bit time_t (2038-01-19T03:14:07 UTC, i.e. INT32_MAX seconds since the epoch). Rejected dates silently fell through icaltime_timegm() as 0 (1970-01-01) instead of the correct value or an explicit error. This corrupts any caller that uses time_t-max as an "unbounded end" sentinel, a common calendar idiom: the sentinel round-trips to epoch, so recurrence/timespan queries silently return truncated results instead of "no upper bound". Confirmed via a real-world case: Debian bug #1142273 (cyrus-imapd FTBFS on i386), traced to exactly this in icalcomponent_foreach_recurrence(). The check has been wrong since it was introduced in 72330c96 (2001-2002); a related commit (ef907061) even called it "arguably a bug" at the time and left it in place. Replace the padded day-level approximation with an exact check against the true boundary (year/month/day/hour/min/sec), and update the two existing regression tests that had encoded the old, incorrect cutoff as expected behavior. Backport of the equivalent fix on the 4.0 branch (libical/libical#1350) to this 3.0 branch, which is what several distributions (e.g. Debian) still package. originally-failing cyrus-imapd cunit test now passes unmodified against a patched i386 libical.so. |
Edmund Lodewijks <edmund@proteamail.com> | no | https://github.com/libical/libical/pull/1351 | 2026-07-26 |