Test Num.Str roundtrippage is stable

Covers R#1651 rakudo/rakudo#1651

Author: zoffixznet

S32-num/stress.t

@@ -0,0 +1,31 @@
+use v6;
+use Test;
+
+plan 1;
+
+# https://github.com/rakudo/rakudo/issues/1651
+subtest 'No drift when roundtripping Num -> Str -> Num -> Str' => {
+    # In this test, it's fine if the original .Str gives the string that
+    # doesn't match what the user has entered (since it may be a number that
+    # doesn't have exact representation in a double). However, .Num.Str
+    # roundtripping *that* string must produce the first string. i.e. there
+    # shouldn't be drift after we figure out what the representable number is
+
+    my @ranges := (
+        (^30 .map: { 10**($_*10)}),  (^30 .map: {-10**($_*10)}),
+        (^30 .map: { 10**($_*-10)}), (^30 .map: {-10**($_*-10)}),
+        3e-324, 3e-320, 3e307
+    );
+
+    plan @ranges * my \iters = 1000;
+    for @ranges -> \r {
+        for ^iters {
+            my \n  := r.rand;
+            my \n1 := n.Str.Num; # get first correct num
+            my \n2 := n.Str.Num.Str.Num.Str.Num.Str.Num; # second
+            cmp-ok n1, '===', n2, "{n} roundtrippage is stable";
+        }
+    }
+}
+
+# vim: ft=perl6