Bug #28601 Incoherent results when using GROUP BY (regression)

Description:
In 5.0.27, the following query gives a different result than 5.0.41

5.0.27
------

mysql> SELECT c1.Continent, City.Name, City.Population FROM Country AS c1 INNER JOIN City ON City.CountryCode = c1.Code WHERE City.Population = (SELECT MAX(City.Population) FROM City INNER JOIN Country ON Country.Code = City.CountryCode WHERE Country.Continent = c1.Continent);
+---------------+-------------------+------------+
| Continent     | Name              | Population |
+---------------+-------------------+------------+
| Oceania       | Sydney            |    3276207 | 
| South America | São Paulo         |    9968485 | 
| Africa        | Cairo             |    6789479 | 
| Asia          | Mumbai (Bombay)   |   10500000 | 
| North America | Ciudad de México  |    8591309 | 
| Europe        | Moscow            |    8389200 | 
+---------------+-------------------+------------+
6 rows in set (1 min 15.86 sec)

5.0.41
------

mysql> SELECT c1.Continent, City.Name, City.Population FROM Country AS c1 INNER JOIN City ON City.CountryCode = c1.Code WHERE City.Population = (SELECT MAX(City.Population) FROM City INNER JOIN Country ON Country.Code = City.CountryCode WHERE Country.Continent = c1.Continent) GROUP BY Continent;
+-----------+-----------------+------------+
| Continent | Name            | Population |
+-----------+-----------------+------------+
| Asia      | Mumbai (Bombay) |   10500000 | 
+-----------+-----------------+------------+
1 row in set (1 min 12.28 sec)

Actually, the GROUP BY clause seems useless, but still, the result should be as previous versions.

How to repeat:
Run the following query on a clean world database.

SELECT c1.Continent, City.Name, City.Population FROM Country AS c1 INNER JOIN City ON City.CountryCode = c1.Code WHERE City.Population = (SELECT MAX(City.Population) FROM City INNER JOIN Country ON Country.Code = City.CountryCode WHERE Country.Continent = c1.Continent);

It might be a duplicate or related to #28337, but not really sure.

Thank you for the report.

I can not repeat described behaviour with current sources, but can with 5.0.41.

As bug #28337 is "Documenting" state now, I don't clssify this one as packaging problem, but will set status to "Can't repeat"