[Issue 20206] New: potential bug in complex power operator

[d-bugmail at puremagic.com]

https://issues.dlang.org/show_bug.cgi?id=20206

          Issue ID: 20206
           Summary: potential bug in complex power operator
           Product: D
           Version: D2
          Hardware: Other
                OS: All
            Status: NEW
          Severity: normal
          Priority: P1
         Component: phobos
          Assignee: nobody at puremagic.com
          Reporter: dlang at croco-puzzle.com

During Phobos PR #7173, all 32bit machines failed in a unittest for the complex
power operator, while all 64bit machines did not. This might be due to a bug in
complex ^^.

The following program will probably reproduce the error, but I had no
opportunity to check this, because I did not succeed in installing dmd on my
old 32bit machine. (Got segmentation fault, whenever I started dmd. I think, a
library is missing. ldc2 complained about not finding gcc, although being
installed and in $PATH. And gdc did not reproduce the error, but used an older
version of phobos.)

void main()
{
    import std.complex;

    auto rec3a = 0.79 ^^ complex(6.8, 5.7);
    auto rec3b = complex(0.79, 0.0) ^^ complex(6.8, 5.7);

    assert(approxEqual(rec3a.re, rec3b.re, double.epsilon));
}

bool approxEqual(T, U, V)(T lhs, U rhs, V maxRelDiff = 1e-9, V maxAbsDiff = 0)
{
    import std.traits:isIntegral;
    import std.math:fabs;

    if (isIntegral!T || isIntegral!U)
    {
        return approxEqual(real(lhs), real(rhs), maxRelDiff, maxAbsDiff);
    }

    if (lhs == rhs) return true;

    static if (is(typeof(lhs.infinity)) && is(typeof(rhs.infinity)))
    {
        if (lhs == lhs.infinity || rhs == rhs.infinity ||
            lhs == -lhs.infinity || rhs == -rhs.infinity) return false;
    }

    auto diff = fabs(lhs - rhs);

    return diff <= maxRelDiff*fabs(lhs) || diff <= maxRelDiff*fabs(rhs) || diff
<= maxAbsDiff;
}

--