`FT_Matrix_Check` doesn't correctly work in 32bit environment
Consider a matrix
{
xx = 655220030,
xy = 982830045,
yx = 39292,
yy = 65536
}
Because the largest value is larger than 23170, FT_Matrix_Check
wants to scale down all values to avoid a 32bit overflow later on. However, while doing so, it performs
scale = FT_DivFix( maxval, 23170 );
(with maxval = 982830045
) which returns -1515047148 in a 32bit environment. In other words, the code to prevent an overflow is overflowing by itself...