Source: postgresql-17
Version: 17.2-1
Severity: serious
Tags: ftbfs
Hi,
The latest tzdata upload, version 2025a-1, causes the postgresql-17 to
FTBFS. The autopkgtest also fails and the log has more details about the
regression:
| 559s + head -100 /tmp/regress.6mRmSV/regression.diffs
| 559s diff -U3
/tmp/autopkgtest-lxc.80ej6hz3/downtmp/build.oIq/src/src/test/regress/expected/timestamptz.out
/tmp/regress.6mRmSV/results/timestamptz.out
| 559s ---
/tmp/autopkgtest-lxc.80ej6hz3/downtmp/build.oIq/src/src/test/regress/expected/timestamptz.out
2025-01-18 06:05:25.253406473 +0000
| 559s +++ /tmp/regress.6mRmSV/results/timestamptz.out 2025-01-18
06:05:29.849506874 +0000
| 559s @@ -2391,7 +2391,7 @@
| 559s SELECT make_timestamptz(1846, 12, 10, 0, 0, 0, 'Asia/Manila') AT TIME
ZONE 'UTC';
| 559s timezone
| 559s --------------------------
| 559s - Wed Dec 09 15:56:00 1846
| 559s + Wed Dec 09 15:56:08 1846
| 559s (1 row)
| 559s
| 559s SELECT make_timestamptz(1881, 12, 10, 0, 0, 0, 'Europe/Paris') AT TIME
ZONE 'UTC';
| 559s + exit 1
| 559s + rm -rf /tmp/regress.6mRmSV
| 559s autopkgtest [06:05:53]: test installcheck: -----------------------]
The full CI log is available there:
https://ci.debian.net/data/autopkgtest/testing/amd64/p/postgresql-17/56552910/log.gz
The testsuite failure is trigger by minor changes of historical data on
the Asia/Manila timezone. More precisely the LMT time used before 6th
September 1899 was refined and adjusted by +8 seconds, causing a
mismatch with postgresql-17 testsuite expectations.
The test should probably be updated, or maybe rather disabled as I am
not fully sure that timezone tests on such old dates really makes sense.
Regards
Aurelien
