Bug#18411494 WRONG COMPARSION ON BIG DECIMAL VALUES

Post-push fix: Avoid negative zero when converting from float to decimal.

Author: Tor Didriksen

mysql-test/r/type_newdecimal.result

@@ -2130,3 +2130,6 @@ SELECT * FROM t1 WHERE value = '100000000000000000000002';
 value
 100000000000000000000002
 DROP TABLE t1;
+SELECT CAST(-0.0e0 AS DECIMAL) = 0;
+CAST(-0.0e0 AS DECIMAL) = 0
+1

mysql-test/t/type_newdecimal.test

@@ -1711,3 +1711,5 @@ DEALLOCATE PREPARE stmt;
 ALTER TABLE t1 ADD INDEX value (value);
 SELECT * FROM t1 WHERE value = '100000000000000000000002';
 DROP TABLE t1;
+
+SELECT CAST(-0.0e0 AS DECIMAL) = 0;

sql/my_decimal.cc

@@ -267,9 +267,6 @@ int str2my_decimal(uint mask, const char *from, size_t length,
     }
   }
   check_result_and_overflow(mask, err, decimal_value);
-  // Avoid returning negative zero, cfr. decimal_cmp()
-  if (decimal_value->sign() && decimal_is_zero(decimal_value))
-    decimal_value->sign(false);
   return err;
 }
 

strings/decimal.c

@@ -996,6 +996,9 @@ internal_str2dec(const char *from, decimal_t *to, char **end, my_bool fixed)
         error= decimal_shift(to, (int) exponent);
     }
   }
+  /* Avoid returning negative zero, cfr. decimal_cmp() */
+  if (to->sign && decimal_is_zero(to))
+    to->sign= FALSE;
   return error;
 
 fatal_error: