JAVA RESULT: -268435456
PERL RESULT: 4026531840

Hope this simple little C program (compiled as C++) clarifies:

// tbs2.cpp
// Built with: g++ -o tbs2 -std=c++11 -Wall -O3 tbs2.cpp (and with 32-
+bit int)
#include <cstdio>
int main(int argc, char* argv[])
{
   printf("d : %d\n",   -268435456);
   printf("u : %u\n",   -268435456);
   printf("x : 0x%x\n", -268435456);
   return 0;
}
[download]

Running this prints:

d : -268435456
u : 4026531840
x : 0xf0000000
[download]

Though all 32 bits are the same (11110000000000000000000000000000), what is displayed depends on whether you interpret them as a signed int (%d) or an unsigned int (%u) or an unsigned int as a hexadecimal number (%x). Java is displaying the 32-bit value as a signed int (-268435456). Perl is displaying the identical value as an unsigned int (4026531840).

Sorry, I haven't used Java for twenty years, but Google tells me it has some sort of printf facility, so displaying your HIGH_BITS with Java printf with "%d" and "%u" - in addition to your System.out.print( HIGH_BITS ) - might be fun.

My original basic test C++ program:

// ANSI C++ 11: tbs1.cpp
// Built with: g++ -o tbs1 -std=c++11 -Wall -O3 tbs1.cpp
#include <cstdint>
#include <iostream>
int main(int argc, char* argv[])
{
   {
      int32_t HIGH_BITS = 0xFFFFFFFF << 28;
      std::cout << "int32       : HIGH_BITS = " << HIGH_BITS << "\n";
   }
   {
      uint32_t HIGH_BITS = 0xFFFFFFFF << 28;
      std::cout << "uint32      : HIGH_BITS = " << HIGH_BITS << "\n";
   }
   {
      int64_t HIGH_BITS = 0xFFFFFFFF << 28;
      std::cout << "int64       : HIGH_BITS = " << HIGH_BITS << "\n";
   }
   {
      uint64_t HIGH_BITS = 0xFFFFFFFF << 28;
      std::cout << "uint64      : HIGH_BITS = " << HIGH_BITS << "\n";
   }
   return 0;
}
[download]

displays:

int32       : HIGH_BITS = -268435456
uint32      : HIGH_BITS = 4026531840
int64       : HIGH_BITS = 4026531840
uint64      : HIGH_BITS = 4026531840
[download]

Further update: For cheap thrills, let's display the bits.

// ANSI C++ 11: tbs4.cpp
// Built with: g++ -o tbs4 -std=c++11 -Wall -O3 tbs4.cpp
#include <cstddef>
#include <cstdint>
#include <cstdio>
#include <iostream>
#include <bitset>
int main(int argc, char* argv[])
{
   printf("0xFFFFFFFF = %u\n", 0xFFFFFFFF);
   printf("0xF0000000 = %u\n", 0xF0000000);
   // This is the OP's 32-bit value HIGH_BITS
   const uint32_t HIGH_BITS = 0xFFFFFFFF << 28;
   std::cout << "HIGH_BITS  = " << HIGH_BITS << "\n";
   // Display all 32 bits             1         2         3
   //                        12345678901234567890123456789012
   const std::bitset<32> b1{"11110000000000000000000000000000"};
   std::cout << "b1         = " << b1 << "\n";
   const std::bitset<32> b2{HIGH_BITS};
   std::cout << "b2         = " << b2 << "\n";
   return 0;
}
[download]

Running this produces:

0xFFFFFFFF = 4294967295
0xF0000000 = 4026531840
HIGH_BITS  = 4026531840
b1         = 11110000000000000000000000000000
b2         = 11110000000000000000000000000000
[download]

64-bit Data Model References

• 64-bit Data Models (wikipedia) - note that the two dominant 64-bit programming models (LP64 and LLP64) both use 32-bit ints
• 64-bit Programming Models: Why LP64? (unix.org)
• What decides the sizeof an integer? (SO)
• Re^4: The 10**21 Problem (Part 4)

Updated: Minor changes were made to the originally posted code.