Message-ID: <gh-pgjdbc-pgjdbc-3930-c3854580552@github.com>
From: "vlsi (@vlsi)" <noreply+vlsi@github.com>
To: "pgjdbc/pgjdbc" <noreply+pgjdbc-pgjdbc@github.com>
Date: Thu, 05 Feb 2026 16:01:23 +0000
Subject: Re: [pgjdbc/pgjdbc] issue #3930: Revert semantic calendar changes introduced with #3837
In-Reply-To: <gh-pgjdbc-pgjdbc-3930@github.com>
References: <gh-pgjdbc-pgjdbc-3930@github.com>
List-Id: <gh-pgjdbc-pgjdbc.github.com>
X-GitHub-Author-Login: vlsi
X-GitHub-Comment-Id: 3854580552
X-GitHub-Comment-Type: issue_comment
X-GitHub-Issue: 3930
X-GitHub-Repo: pgjdbc/pgjdbc
X-GitHub-Type: comment
X-GitHub-Url: https://github.com/pgjdbc/pgjdbc/issues/3930#issuecomment-3854580552
Content-Type: text/plain; charset=utf-8

Looks like we should indeed better keep backward compatibility rather than introduce incompatible chainges in patch releases.

```
  The Current Issue

  The fix introduced a semantic breaking change that causes apparent "data corruption" for historical dates (pre-1582):
  ┌────────────────────────────────────────────────┬──────────────────┐
  │                    Scenario                    │      Result      │
  ├────────────────────────────────────────────────┼──────────────────┤
  │ Write 1000-01-02 with 42.7.9, read with 42.7.8 │ Shows 1000-01-07 │
  ├────────────────────────────────────────────────┼──────────────────┤
  │ Write 1000-01-02 with 42.7.8, read with 42.7.9 │ Shows 999-12-28  │
  └────────────────────────────────────────────────┴──────────────────┘


  Possible Resolutions
  ┌──────────────────────────┬───────────────────────────────────────┬──────────────────────────────────────────┐
  │          Option          │                 Pros                  │                   Cons                   │
  ├──────────────────────────┼───────────────────────────────────────┼──────────────────────────────────────────┤
  │ Full revert              │ Restores backward compatibility       │ Keeps the technically incorrect behavior │
  ├──────────────────────────┼───────────────────────────────────────┼──────────────────────────────────────────┤
  │ Major version (43.0.0)   │ Proper semver for breaking change     │ Requires re-release                      │
  ├──────────────────────────┼───────────────────────────────────────┼──────────────────────────────────────────┤
  │ Connection property      │ User choice, both behaviors available │ Added complexity, not clear default      │
  ├──────────────────────────┼───────────────────────────────────────┼──────────────────────────────────────────┤
  │ Default old + opt-in new │ Backward compatible with path forward │ Technical correctness not default        │
  └──────────────────────────┴───────────────────────────────────────┴──────────────────────────────────────────┘
```

I do not see win-win case here :-/

Frankly, I am leaning towards "revert behavior, and add an option that could match driver's behavior with database behavior".