Re: op/numconvert.t failures

[p5sagit/p5-mst-13.2.git] / t / op / numconvert.t

diff --git a/t/op/numconvert.t b/t/op/numconvert.t
index d41594e..084092e 100755 (executable)
--- a/t/op/numconvert.t
+++ b/t/op/numconvert.t
@@ -224,9 +224,15 @@ for my $num_chain (1..$max_chain) {
                     and $ans[0] == $ans[1] and $ans[0] <= ~0
                      # First must be in E notation (ie not just digits) and
                      # second must still be an integer.
+                    # eg 1.84467440737095516e+19
+                    # 1.84467440737095516e+19 for 64 bit mantissa is in the
+                    # integer range, so 1.84467440737095516e+19 + 0 is treated
+                    # as integer addition. [should it be?]
+                    # and 18446744073709551600 + 0 is 18446744073709551600
+                    # Which isn't the string you first thought of.
                      # I can't remember why there isn't symmetry in this
                      # exception, ie why only the first ops are tested for 'N'
-                     and $ans[0] !~ /^-?\d+$/ and $ans[0] !~ /^-?\d+$/) {
+                     and $ans[0] != /^-?\d+$/ and $ans[1] !~ /^-?\d+$/) {
              print "# ok, numerically equal - notation changed due to adding zero\n";
            } else {
              $nok++,