use always 64-bit arithemtics

long is only 32-bit on 32-bit systems and on 64-bit Windows
ensure we have consistent behavior

src/CglGMI/CglGMI.cpp

@@ -884,11 +884,11 @@ bool CglGMI::scaleCut(double* cutElem, int* cutIndex, int cutNz,
 /************************************************************************/
 bool CglGMI::scaleCutIntegral(double* cutElem, int* cutIndex, int cutNz,
 			      double& cutRhs) {
-  long gcd, lcm;
+  int64_t gcd, lcm;
   double maxdelta = param.getEPS(); 
   double maxscale = 1000; 
-  long maxdnom = 1000; 
-  long numerator = 0, denominator = 0;
+  int64_t maxdnom = 1000;
+  int64_t numerator = 0, denominator = 0;
   // Initialize gcd and lcm
   CoinRational r = CoinRational(cutRhs, maxdelta, maxdnom);
   if (r.getNumerator() != 0){
@@ -933,13 +933,13 @@ bool CglGMI::scaleCutIntegral(double* cutElem, int* cutIndex, int cutNz,
 } /* scaleCutIntegral */
 
 /************************************************************************/
-long CglGMI::computeGcd(long a, long b) {
+int64_t CglGMI::computeGcd(int64_t a, int64_t b) {
   // This is the standard Euclidean algorithm for gcd
-  long remainder = 1;
+  int64_t remainder = 1;
   // Make sure a<=b (will always remain so)
   if (a > b) {
     // Swap a and b
-    long temp = a;
+    int64_t temp = a;
     a = b;
     b = temp;
   }

src/CglGMI/CglGMI.hpp

@@ -247,11 +247,11 @@ class CGLLIB_EXPORT CglGMI : public CglCutGenerator {
 			  double& cutRhs);
 
   /// Compute the nearest rational number; used by scale_row_integral
-  bool nearestRational(double val, double maxdelta, long maxdnom,
-			long& numerator, long& denominator);
+  bool nearestRational(double val, double maxdelta, int64_t maxdnom,
+			int64_t& numerator, int64_t& denominator);
 
   /// Compute the greatest common divisor
-  long computeGcd(long a, long b);
+  int64_t computeGcd(int64_t a, int64_t b);
 
   /// print a vector of integers
   void printvecINT(const char *vecstr, const int *x, int n) const;

src/CglGomory/CglGomory.cpp

@@ -419,13 +419,13 @@ static long long int gcd(long long int a, long long int b)
 #define GOMORY_INT int
 #endif
 #if USE_CGL_RATIONAL>0
-static long computeGcd(long a, long b) {
+static int64_t computeGcd(int64_t a, int64_t b) {
   // This is the standard Euclidean algorithm for gcd
-  long remainder = 1;
+  int64_t remainder = 1;
   // Make sure a<=b (will always remain so)
   if (a > b) {
     // Swap a and b
-    long temp = a;
+    int64_t temp = a;
     a = b;
     b = temp;
   }
@@ -448,10 +448,10 @@ static long computeGcd(long a, long b) {
 } /* computeGcd */
 static bool scaleCutIntegral(double* cutElem, int* cutIndex, int cutNz,
 			     double& cutRhs, double maxdelta) {
-  long gcd, lcm;
+  int64_t gcd, lcm;
   double maxscale = 1000; 
-  long maxdnom = USE_CGL_RATIONAL;
-  //long numerator = 0, denominator = 0;
+  int64_t maxdnom = USE_CGL_RATIONAL;
+  //int64_t numerator = 0, denominator = 0;
   // Initialize gcd and lcm
   CoinRational r = CoinRational(cutRhs, maxdelta, maxdnom);
   if (r.getNumerator() != 0){

src/CglPreProcess/CglPreProcess.cpp

@@ -1183,13 +1183,13 @@ static void writeDebugMps(const OsiSolverInterface *solver,
 #define USE_CGL_RATIONAL 1
 #if USE_CGL_RATIONAL>0
 #include "CoinRational.hpp"
-static long computeGcd(long a, long b) {
+static int64_t computeGcd(int64_t a, int64_t b) {
   // This is the standard Euclidean algorithm for gcd
-  long remainder = 1;
+  int64_t remainder = 1;
   // Make sure a<=b (will always remain so)
   if (a > b) {
     // Swap a and b
-    long temp = a;
+    int64_t temp = a;
     a = b;
     b = temp;
   }
@@ -1212,15 +1212,15 @@ static long computeGcd(long a, long b) {
 } /* computeGcd */
 static bool scaleRowIntegral(double* rowElem, int rowNz)
 {
-  long gcd, lcm;
+  int64_t gcd, lcm;
   double maxdelta = 1.0e-13;
   double maxscale = 1000; 
-  long maxdnom = 1000;
-  //long numerator = 0, denominator = 0;
+  int64_t maxdnom = 1000;
+  //int64_t numerator = 0, denominator = 0;
   // Initialize gcd and lcm
   CoinRational r = CoinRational(rowElem[0], maxdelta, maxdnom);
   if (r.getNumerator() != 0){
-    gcd = labs(r.getNumerator());
+    gcd = llabs(r.getNumerator());
     lcm = r.getDenominator();
   } else {
     return false;