trying pynac PR 379

build/pkgs/pynac/patches/379.diff

@@ -0,0 +1,48 @@
+diff --git a/ginac/numeric.cpp b/ginac/numeric.cpp
+index 512a6c8..2206044 100644
+--- a/ginac/numeric.cpp
++++ b/ginac/numeric.cpp
+@@ -1414,6 +1414,25 @@ const numeric numeric::sub(const numeric &other) const {
+         }
+ }
+
++#if defined __has_builtin
++#  if __has_builtin (__builtin_smull_overflow)
++#    define smull_overflow __builtin_smull_overflow
++#  endif
++#endif
++
++#if !defined smull_overflow
++static int smull_overflow(long a, long b, long *result) {
++        static long lsqrt = std::lround(std::sqrt(std::numeric_limits<long>::max()));
++        if (-lsqrt < a and a < lsqrt and -lsqrt < b and b < lsqrt) {
++                // Will not overflow.
++                *result = a * b;
++                return 0;
++        }
++        // May overflow.
++        return 1;
++}
++#endif
++
+ /** Numerical multiplication method.  Multiplies *this and argument and returns
+  *  result as a numeric object. */
+ const numeric numeric::mul(const numeric &other) const {
+@@ -1433,7 +1452,7 @@ const numeric numeric::mul(const numeric &other) const {
+         switch (t) {
+         case LONG: {
+                 long result;
+-                if (!__builtin_smull_overflow(v._long, other.v._long, & result))
++                if (!smull_overflow(v._long, other.v._long, & result))
+                         return result;
+                 // the multiplication overflowed, so use mpz
+                 mpz_t bigint;
+@@ -2319,7 +2338,7 @@ numeric & operator*=(numeric & lh, const numeric & rh)
+         switch (lh.t) {
+         case LONG: {
+                 long result;
+-                if (!__builtin_smull_overflow(lh.v._long, rh.v._long, & result)) {
++                if (!smull_overflow(lh.v._long, rh.v._long, & result)) {
+                         lh.v._long = result;
+                         lh.hash = (lh.v._long==-1) ? -2 : lh.v._long;
+                         return lh;