I don`t know what bug is in my programm. Who can give me this test.

As I remember, in 12th test the point is in precision. If you write in Pascal, use Extended (in C++ - double), use very accurate value for PI (3.1415926 is NOT enough), in Pascal use standard constant "pi", in C - 2.0*acos(0.0)

Well, in my case the reason for WA#12 was inproper implementation of Graham's algorithm, not a precision. Be careful with cases, when there are points with the same polar angle.