It contains the following code:

  if (AbsFitsIntoUint64(bigint)) {
    double absolute_value = static_cast<double>(AbsToUint64(bigint));
    double result = bigint.IsNegative() ? -absolute_value : absolute_value;
    return Double::New(result);
  }

all previously supported systems agreed on what they do when compiling static_cast<double>(uint64_value) however MSVC 2010 on Windows 64 does not agree with other systems. Here is a demo:

if defined(_MSC_VER)

include <stdint.h>

define PRIu64 "I64u"

else

ifndef __STDC_FORMAT_MACROS

define __STDC_FORMAT_MACROS

endif

include <inttypes.h>

endif

include <cstdio>

double convert(uint64_t x) {
  return static_cast<double>(x);
}

uint64_t bitcast(double x) {
  return (uint64_t)&x;
}

int main(int argc, char* argv[]) {
  const uint64_t a = 12345678901234567891ULL;
  const double b = convert(a);
  const uint64_t c = bitcast(b);
  const uint64_t d = 4892433759222981601ULL;

  printf("{%" PRIu64 " => %lf (%" PRIu64 ")} %s %" PRIu64 "\n",
         a, b, c, c == d ? "==" : "!=", d);

  return 0;
}

∮ g++ -m32 -o test test.cc
∮ ./test
{12345678901234567891 => 12345678901234567168.000000 (4892433759222981601)} == 4892433759222981601
∮ g++ -m64 -o test test.cc
∮ ./test
{12345678901234567891 => 12345678901234567168.000000 (4892433759222981601)} == 4892433759222981601

On MSVC x86
D:\src\temp3>cl /nologo test.cc
D:\src\temp3>test.exe
{12345678901234567891 => 12345678901234567000.000000� (4892433759222981601)} == 4892433759222981601

On MSVC x64
D:\src\temp3>cl /nologo test.cc
D:\src\temp3>test.exe
{12345678901234567891 => 12345678901234569000.000000 (4892433759222981602)} != 4892433759222981601

This difference causes corelib/big_integer_vm_test to fail on Windows x64. I am going to mark this failure as expected for now but we need to have a work-around for this incompatibility.