On 05.08.25 20:06, Peter Eisentraut wrote:
> diff --git a/src/backend/utils/adt/timestamp.c b/src/backend/utils/
adt/timestamp.c
> index 25cff56c3d0..e640b48205b 100644
> --- a/src/backend/utils/adt/timestamp.c
> +++ b/src/backend/utils/adt/timestamp.c
> @@ -4954,7 +4954,7 @@ timestamptz_trunc_internal(text *units,
TimestampTz timestamp, pg_tz *tzp)
> case DTK_SECOND:
> case DTK_MILLISEC:
> case DTK_MICROSEC:
> - PG_RETURN_TIMESTAMPTZ(timestamp);
> + return timestamp;
> break;
>
> default:
This one is an actual bug. With int64-pass-by-reference,
PG_RETURN_TIMESTAMPTZ() returns a newly allocated address. This is
easily reproduced by
SELECT date_trunc( 'week', timestamp with time zone 'infinity' );
which will return a random value on 32-bit systems. (It correctly
returns 'infinity' otherwise.)
I'll see about backpatching fixes for this.
This is new in PG18, so I'm continuing the discussion in the original
thread:
https://www.postgresql.org/message-id/flat/CAAvxfHc4084dGzEJR0_pBZkDuqbPGc5wn7gK_M0XR_kRiCdUJQ%40mail.gmail.com
