To reproduce the bug, use this script. Let 'TestDatabase' be any kind of database, and let the table field 'Table'.'NullField' be any database field that contains a NULL value.

<?php
try {
$DB = new COM("ADODB.Connection") or die("Cannot start ADO");
$DB->Open("DRIVER={SQL Server};SERVER=SERVER\WEB;DATABASE=TestDatabase");

$RS = new COM("ADODB.Recordset");
$RS->Open("SELECT TEXT FROM Table", $DB);

echo $RS["NullField"];

} catch (Exception $e) {
	echo $e->getTraceAsString() . "<br>";
	echo $e->getMessage();
}
?>

Could you please provide an example that doesn't require any external resources (especially those that exist only under a particular OS) ?
Also please post what you expect and what you got when you execute the reproduce code. 

The only way I am able to reproduce this behavior is using COM, but the nature of the bug is not directly linked to the COM libraries itself.
Because of this, I am unable to make a test case that is platform independent and that does not require external resources.

I will try just once more:
Create a Visual Basic ActiveX DLL project. Name the project 'NullClass', and name the class 'Null'. Add this single function to the class code:

Public Property Get Value()
    Value = Null
End Property

Place the cursor inside the function, Click the menu "Tools", "Procedure Properties", "Advanced", and select "User Interface Default". Click "File", "Make NullClass.dll".

Then run this PHP script:

<?php
$Obj = new COM('NullClass.Null');
echo $Obj;
?>

EXPECTED: The script does not output anything, because the object contains a NULL value.

ACTUAL: The script either outputs garbage or causes an Access Violation message.

Please try using this CVS snapshot:

  http://snaps.php.net/php5-latest.tar.gz
 
For Windows:
 
  http://snaps.php.net/win32/php5-win32-latest.zip

There seems to be no change in the latest (5.1.x) snapshot, compiled 1 march 2005 7:55 GMT.

Please provide a patch or at least reproduce code that doesn't require Windoze. 

ZEND_API void zend_make_printable_zval(zval *expr, zval *expr_copy, int *use_copy)
{
...
...
case IS_OBJECT:
{
...
...
if (Z_OBJ_HANDLER_P(expr, cast_object)) {
if( (Z_OBJ_HANDLER_P(expr, cast_object)(expr, expr_copy, IS_STRING, 0 TSRMLS_CC) == SUCCESS) &&
   (expr_copy->type == IS_STRING)) {
		break;
}
}

You're still pasting some C source code. Can you please explain what you want to say? Is that last comment a patch or reproduce code? (never heard of 'diff' utility?)

Then just forget it, if you cannot just read my first post and just go into the code and see what is there.

It's bogus, since you can't explain what are you talking about.