Added --compact to mysqlbinlog

Fixed output from mysqlbinlog when using --skip-comments
Fixed warnings from valgrind
Fixed ref_length when used with HEAP tables
More efficent need_conversion()
Fixed error handling in UPDATE with not updateable tables
Fixed bug in null handling in CAST to signed/unsigned



client/client_priv.h:
  cleanup & added OPT_COMPACT
client/mysqldump.c:
  Added option --compact to get a compact readable dump.
  Ensure that SET CHARACTER_SET_CLIENT is not done if we have not remembered the old character set
  Print optimization comments even if --skip-comments are given as these are not true comments. (Before these where only printed at end, which was a bug)
mysql-test/r/cast.result:
  More cast tests
mysql-test/r/derived.result:
  Removed warnings
mysql-test/r/mysqldump.result:
  Update results after fixing mysqlbinlog
mysql-test/r/query_cache.result:
  Make test usable with --extern
  more tests
mysql-test/r/rpl_until.result:
  Make test repeatable under valgrind
mysql-test/r/sql_mode.result:
  Fix test result
mysql-test/r/subselect.result:
  Make test smaller. Update wrong results
mysql-test/t/cast.test:
  More cast tests
mysql-test/t/derived.test:
  Removed warnings
mysql-test/t/query_cache.test:
  Make test usable with --extern
  more tests
mysql-test/t/rpl_until.test:
  fix for valgrind.  Becasue of unknown reason one got 'Slave_SQL_Running=yes' in this setup
mysql-test/t/subselect.test:
  Make test case smaller
sql/field.cc:
  Updated need_conversion() to use new arguments
sql/ha_heap.cc:
  Moved initialization of ref_length to right place. This fixed problem that we had a ref_length of 8 for heap tables, which was not efficent.
sql/item_func.cc:
  Cleanup
sql/item_func.h:
  Fixed bug in null_handling for cast to signed/unsigned
sql/item_strfunc.cc:
  Optimized/cleaned up Item_func_conv_charset3
sql/item_sum.cc:
  Cleanup.
  Ensure that some flag variables are cleared in cleanup()
sql/item_sum.h:
  Fixed references to uninitialized memory
sql/opt_range.cc:
  Fixed spelling error
sql/sql_class.cc:
  Fixed wrong return code, which could case protocol problems
sql/sql_class.h:
  After merge fix
sql/sql_prepare.cc:
  Added comments
sql/sql_show.cc:
  Cleanup
sql/sql_string.cc:
  Optimzed usage of need_conversion().
  - Removed not used argument
  - Save diff lenght in 'offset' to not have to recalculate length several times.
  Cleaned up comment
  Optimized copy_aligned() based on the knowledge that it's only called when you have wrong data
sql/sql_string.h:
  Updated need_conversion() and copy_aligned() to use new arguments
sql/sql_update.cc:
  Fixed error handling with non-updateable tables
sql/sql_yacc.yy:
  Ensure that lex->lock_options are set correctly (to get rid of warnings from valgrind)
  Ensure that cast_type sets lex->charset and lex->length. Without these CONVERT() didn't work properly

sql/sql_string.h

@@ -183,9 +183,11 @@ public:
   bool copy();					// Alloc string if not alloced
   bool copy(const String &s);			// Allocate new string
   bool copy(const char *s,uint32 arg_length, CHARSET_INFO *cs);	// Allocate new string
-  static bool needs_conversion(const char *s, uint32 arg_length,
-  			       CHARSET_INFO *cs_from, CHARSET_INFO *cs_to);
-  bool copy_aligned(const char *s, uint32 arg_length, CHARSET_INFO *cs);
+  static bool needs_conversion(uint32 arg_length,
+  			       CHARSET_INFO *cs_from, CHARSET_INFO *cs_to,
+			       uint32 *offset);
+  bool copy_aligned(const char *s, uint32 arg_length, uint32 offset,
+		    CHARSET_INFO *cs);
   bool set_or_copy_aligned(const char *s, uint32 arg_length, CHARSET_INFO *cs);
   bool copy(const char*s,uint32 arg_length, CHARSET_INFO *csfrom,
 	    CHARSET_INFO *csto);