Possible bug with row_to_json

When using a subquery as a source for row_to_json, depending on the
order of arguments it may ignore renaming a column.

jack=# create table player(
jack(# player_id serial primary key,
jack(# name varchar not null unique
jack(# );
NOTICE: CREATE TABLE will create implicit sequence
"player_player_id_seq" for serial column "player.player_id"
NOTICE: CREATE TABLE / PRIMARY KEY will create implicit index
"player_pkey" for table "player"
NOTICE: CREATE TABLE / UNIQUE will create implicit index
"player_name_key" for table "player"
CREATE TABLE
jack=# insert into player(name) values('Jack');
INSERT 0 1
jack=# select row_to_json(t)
jack-# from (
jack(# select player_id as renamed, name
jack(# from player
jack(# order by name
jack(# ) t;
row_to_json
-------------------------------
{"player_id":1,"name":"Jack"}
(1 row)

It ignored the rename.

jack=# select row_to_json(t)
from (
select name, player_id as renamed
from player
order by name
) t;
row_to_json
-----------------------------
{"name":"Jack","renamed":1}
(1 row)

But here it didn't.

Is this a bug?

Jack Christensen