Bruno Haible via GNU coreutils General Discussion <[email protected]>
writes:
> The CI reports that the date/date-tz test fails on FreeBSD, NetBSD, OpenBSD.
>
> On FreeBSD:
>
> FAIL: tests/date/date-tz
> ========================
>
> Mon Jun 15 20:17:00 UTC 2026
> --- exp 2026-06-15 21:48:53.793131000 +0000
> +++ out 2026-06-15 21:48:53.793024000 +0000
> @@ -1 +1 @@
> -2024-10-27T02:30:00+0100
> +2024-10-27T02:30:00+0200
> FAIL tests/date/date-tz.sh (exit status: 1)
That is a recently added test here:
# An ambiguous local time in the fall-back overlap takes the earlier,
# still-DST offset (+0100 here, not +0200).
TZ=Europe/Paris date -d '2024-10-27 02:30:00' '+%Y-%m-%dT%T%z' > out ||
fail=1
printf "2024-10-27T02:30:00+0100\n" > exp || framework_failure_
compare exp out || fail=1
It looks like %z behaves strangely on BSD. Using the following test
program:
$ cat main.c
#include <time.h>
#include <stdio.h>
#include <stdlib.h>
int
main (void)
{
tzset();
time_t t = 1729992600;
struct tm tm;
gmtime_r (&t, &tm);
char buf[256];
strftime (buf, sizeof buf, "%z", &tm);
printf ("tm_gmtoff: %d\n", tm.tm_gmtoff);
printf ("tm_isdst: %d\n", tm.tm_isdst);
puts (buf);
return EXIT_SUCCESS;
}
On GNU/Linux:
$ gcc main.c; TZ=Europe/Paris ./a.out
tm_gmtoff: 0
tm_isdst: 0
+0000
On NetBSD:
$ gcc main.c; TZ=Europe/Paris ./a.out
tm_gmtoff: 0
tm_isdst: 0
+0100
Paul, what do you think? This feels like a bug in BSD to me, but I
certainly trust your opinion here more than mine.
Collin
