Being not able to fix the rounding error in a solution is not a kata issue.

Expected: 702912415374435841
But was: 702912415374435840
How am I supposed to fix the rounding error?

Nice excercise

Could you link to an example that is passing despite not being "float proof"? The top C# answer by g964 seems to be float proof. It correctly returns -1 for 4503599627370497 and correctly returns 67108865 for custom test case 4503599761588225 (which is greater than floating point precision 2^52 = 4503599627370496) (I did my math through wolframalpha which is more precise than Google's calculator).

C# does not lack precision to solve this test, and any problems with precision are introduced by a buggy solution, not by tests.

For 4503599627370497, expeced answer is -1, because it is not a perfect square (but 4503599627370496 is).
For 4503599627370495, expeced answer is -1, because it is not a perfect square (but 4503599627370496 is).

Solutions which cannot handle these inputs are buggy and need to be fixed, it is not a problem with tests.

Running into something similar for the test case of 4503599627370495

The tests in C# are way beyond 7 Kyu

I very much agree, this is way harder than any other 7 Kyu I've ever seen

4503599627370497 is too large a number for double-precision math, meaning it gets rounded down to 4503599627370496 before the square root is found, which is a perfect square. This is way beyond a 7 Kyu exercise

In such case, 67,108,864 ** 2 should end with the digit 6, wouldn't it?

I used https://www.gigacalculator.com/calculators/square-root-calculator.php and calculator app by Google. Both gave me a whole number which was 67,108,864. The google calculator gave me specifically 67,108,864.000000

I am not sure how you checked it, but square root of 4503599627370497 is not a whole number.

Anyone know what is wrong with the arugment being 4503599627370497? I checked and I got a whole number as the square root yet its telling me that I need to return -1.

this was a good one :)