BUG #19739: Parameterized first autocommit statement can have different `transaction_timestamp()` and `statement - Mailing list pgsql-bugs

From PG Bug reporting form
Subject BUG #19739: Parameterized first autocommit statement can have different `transaction_timestamp()` and `statement
Date
Msg-id 19739-bdca3e8fd9b496fa@postgresql.org
Whole thread
Responses Re: BUG #19739: Parameterized first autocommit statement can have different `transaction_timestamp()` and `statement
List pgsql-bugs
The following bug has been logged on the website:

Bug reference:      19739
Logged by:          Shallow
Email address:      theshallow27@gmail.com
PostgreSQL version: 18.6
Operating system:   Linux
Description:

The documentation says `transaction_timestamp()` and
`statement_timestamp()` return the same value during the first statement of
a
transaction. With a parameterized query executed as the first command on a
fresh autocommit connection, they differ. The same query without parameters
returns equality. Parameterized execution uses the extended query protocol,
so
this may expose a distinction between Parse/Bind/Execute message timing and
the documented “first statement” guarantee.

**Reproduction:** Using Psycopg with autocommit enabled on a fresh
connection:

```python
cur.execute(
    "SELECT transaction_timestamp() = statement_timestamp() WHERE %s::int =
0",
    (0,),
)
print(cur.fetchone())  # (False,)
```

Control query on a fresh autocommit connection:

```sql
SELECT transaction_timestamp() = statement_timestamp(); -- true
```

**Expected result:** The parameterized query, as the first SQL statement in
an
implicit transaction, should return `true`; alternatively, the documentation
should clarify how extended-protocol messages affect the guarantee.





pgsql-bugs by date:

Previous
From: PG Bug reporting form
Date:
Subject: BUG #19738: `uuidv7(interval)` rejects sub-millisecond timestamps within the final 48-bit millisecond
Next
From: PG Bug reporting form
Date:
Subject: BUG #19740: `has_language_privilege` returns TRUE for a nonexistent language OID when the user is a superuser