arith_uint256 minus operation differs from CBigNum #11088

Using the 0.8 codebase compiled with mingw and the 0.13 codebase cross compiled for Windows on Ubuntu. With some bespoke code the same values for a minus operation differ between arith_uint256 and CBigNum, I'm getting positive arith_uint256 values where as with CBigNum in older codebases returns negative. I assume that the arith_uint256 is wrapping around, CBigNum being a uint should not have gone negative.

0.8

CBigNum mathsInputOne;
CBigNum mathsInputTwo;
CBigNum mathsOne;

0.13

arith_uint256 mathsInputOne;
arith_uint256 mathsInputTwo;
arith_uint256 mathsOne;

The operation is a simple one.

mathsOne = mathsInputTwo - mathsInputOne;

Value for mathsInputOne = 00000ec639000000000000000000000000000000000000000000000000000000 Value for mathsInputTwo = 00000ec302000000000000000000000000000000000000000000000000000000

The results are below.

0.8 CBigNum (Converted to hexidecimal)

-337000000000000000000000000000000000000000000000000000000

0.13 arith_uint256

fffffffcc9000000000000000000000000000000000000000000000000000000

CBigNum used OpenSSL BN_sub for minus operations where as arith_uint256 is implemented locally in arith_uint256.h. Unfortunately the project I'm working with expects the negative values rather than the wrapped ones, not exactly sure what to do except for restore CBigNum along with OpenSSL bn.h.

It's a uint. Our behavior is correct. If you need signed numbers, you need something else-- or rework your arithmetic to not need signed numbers.

Thanks. I've added CBigNum into the project to get it working as it use to.