DBMS bug in Oracle 10g Relese 2

Question

It seems some bug appeared in Oracle 10g Release 2.
We have a single installation of that release on a Linux system.
Therefore we don't know if the bug also exists on Windows or Mac platforms.
It will be appreciated if someone can give some advice or test the issue on windows.
Here are the steps of reproducing the bug..
-- 1. Create a table with statement.
create table spreaded
( grp NUMBER,
p1 NUMBER,
p2 NUMBER,
p3 NUMBER,
p4 NUMBER
--2. Populate the table with the following PL/SQL block (or otherwise)
declare
begin
for k in 1..400000 loop -- depends on installation's memory parameters
insert into spreaded values (k, 1, NULL, NULL, NULL);
insert into spreaded values (k, NULL, 2, NULL, NULL);
insert into spreaded values (k, NULL, NULL, 3, NULL);
insert into spreaded values (k, NULL, NULL, NULL, 4);
end loop;
end;
--3. Execute the following query
select grp,
count(grp) fake,
avg(p1) prop_1,
avg(p2) prop_2,
avg(p3) prop_3,
avg(p4) prop_4
from spreaded
group by (grp) order by fake;
-- The expected first line of correct result is
GRP FAKE PROP_1 PROP_2 PROP_3 PROP_4
*** 4 1 2 3 4
where *** is some integer
-- The buggy case shows
GRP FAKE PROP_1 PROP_2 PROP_3 PROP_4
*** 1 4
FAKE is not 4 and there are NULL values.
This is against the way of handling nulls by aggregate functions.
If you get correct results, please, add new lines to table through executing PL/SQL
block with modified loop (for k in 400001..800000 loop) to double the row count and then execute select statement from 3.

Answer 1

Some bugs appears, some others go away, but if you think have found a bug, raise a support request to the metalink website.
Nicolas.