Fix for bug #31236: Inconsistent division by zero behavior for

                    floating point numbers

Some math functions did not check if the result is a valid number
(i.e. neither of +-inf or nan).

Fixed by validating the result where necessary and returning NULL in
case of invalid result.


BitKeeper/deleted/.del-matherr.c:
  Rename: sql/matherr.c -> BitKeeper/deleted/.del-matherr.c
configure.in:
  Removed DONT_USE_FINITE, it is not used anywhere.
include/my_global.h:
  isfinite() is a C99 macro which absoletes finite(). First try to use
  it, then fall back to finite() if the target platform has it,
  otherwise use our own implementation.
mysql-test/r/func_math.result:
  Added a test case for bug #31236.
mysql-test/r/strict.result:
  Fixed a test case which relied on old behavior.
mysql-test/t/func_math.test:
  Added a test case for bug #31236.
mysql-test/t/strict.test:
  Fixed a test case which relied on old behavior.
sql/field.cc:
  No need to check if the finite() or its equivalent is available.
sql/item_func.cc:
  Use fix_result() wherever the result can be one of +-inf or nan, 
  assuming the function arguments are valid numbers.
  Removed fix_result() from functions that are defined for all possible
  input numbers.
sql/item_func.h:
  Moved fix_result() from Item_dec_func to Item_func which is a common
  ancestor for Item_dec_func and Item_num_op.
sql/unireg.h:
  Removed POSTFIX_ERROR because no code returns it.

1 files changed, 7 insertions, 12 deletions

diff --git a/sql/item_func.h b/sql/item_func.h
index e09b584de95..cdcbbdab150 100644
--- a/sql/item_func.h
+++ b/sql/item_func.h
@@ -192,6 +192,13 @@ public:
                      void * arg, traverse_order order);
   bool is_expensive_processor(uchar *arg);
   virtual bool is_expensive() { return 0; }
+  inline double fix_result(double value)
+  {
+    if (isfinite(value))
+      return value;
+    null_value=1;
+    return 0.0;
+  }
 };
 
 
@@ -499,18 +506,6 @@ class Item_dec_func :public Item_real_func
     decimals=NOT_FIXED_DEC; max_length=float_length(decimals);
     maybe_null=1;
   }
-  inline double fix_result(double value)
-  {
-#ifndef HAVE_FINITE
-    return value;
-#else
-    /* The following should be safe, even if we compare doubles */
-    if (finite(value) && value != POSTFIX_ERROR)
-      return value;
-    null_value=1;
-    return 0.0;
-#endif
-  }
 };
 
 class Item_func_exp :public Item_dec_func