On Monday, May 9, 2022, Tom Lane <t...@sss.pgh.pa.us> wrote: > =?UTF-8?B?5q2j5Y2O5ZCV?= <kain...@gmail.com> writes: > > I test the following SQL in pg15dev (seems same behavior as the > > previous version). > > select '2020701'::date; > > date > > ------------ > > 0202-07-01 > > (1 row) > > At the first glance, the result seems quite strange. > > [ shrug... ] You left out a zero. It's not apparent to me that > this answer is wrong.
> If you want to pursue this question, you could start by bisecting > to find just which commit changed it and why. > > Manual history inspection of datetime.c https://github.com/postgres/postgres/commit/7778ddc7a2d5b006edbfa69cdb44b8d8c24ec1ff https://www.postgresql.org/message-id/flat/8977CB36860C5843884E0A18D8747B0372BC6401%40szxeml558-mbs.china.huawei.com#8aa7c8b8916056a53f6c6dff0f0eb428 The goal seemed to be able to accept 5-digit years…this behavior change didn’t show in the tests (or discussion) though I didn’t look for the of testing the pre-existing failure mode. David J.
