sin, cos, tan, etc all use radians as arguments and the sin of 45 radians is .859035247...

These are double precision functions, and give about 16 correct significant decimal digits. So floating point roundoff for these functions would not result in an error as large as you guessed.

Regards,

Dave