- MDEV-6695 Bad column name for UCS2 string literals

  The Item_string constructors called set_name() on the source string,
  which was wrong because in case of UCS2/UTF16/UTF32 the source value
  might be a not well formed string (e.g. have incomplete leftmost character).
  Now set_name() is called on str_value after its copied 
  (with optionally left zero padding) from the source string.
- MDEV-6694 Illegal mix of collation with a PS parameter
  Item_param::convert_str_value() did not set repertoire.
  Introducing a new structure MY_STRING_METADATA to collect
  character length and repertoire of a string in a single loop,
  to avoid two separate loops. Adding a new class Item_basic_value::Metadata
  as a convenience wrapper around MY_STRING_METADATA, to reuse the
  code between Item_string and Item_param.

include/m_ctype.h

@@ -735,6 +735,14 @@ my_bool my_propagate_simple(CHARSET_INFO *cs, const uchar *str, size_t len);
 my_bool my_propagate_complex(CHARSET_INFO *cs, const uchar *str, size_t len);
 
 
+typedef struct 
+{
+  size_t char_length;
+  uint repertoire;
+} MY_STRING_METADATA;
+
+void my_string_metadata_get(MY_STRING_METADATA *metadata,
+                            CHARSET_INFO *cs, const char *str, size_t len);
 uint my_string_repertoire(CHARSET_INFO *cs, const char *str, ulong len);
 my_bool my_charset_is_ascii_based(CHARSET_INFO *cs);
 my_bool my_charset_is_8bit_pure_ascii(CHARSET_INFO *cs);

mysql-test/r/ctype_ucs.result

@@ -5333,5 +5333,12 @@ SELECT CONCAT(CONVERT('pi=' USING ucs2),PI()) AS PI;
 PI
 pi=3.141593
 #
+# MDEV-6695 Bad column name for UCS2 string literals
+#
+SET NAMES utf8, character_set_connection=ucs2;
+SELECT 'a','aa';
+a	aa
+a	aa
+#
 # End of 10.0 tests
 #

mysql-test/r/ctype_utf8.result

@@ -6008,5 +6008,28 @@ CONCAT(a, IF(b>10, _utf8 X'61', _utf8 B'01100001'))
 aa
 DROP TABLE t1;
 #
+#  MDEV-6694 Illegal mix of collation with a PS parameter
+#
+SET NAMES utf8;
+CREATE TABLE t1 (a INT, b VARCHAR(10) CHARACTER SET latin1);
+INSERT INTO t1 VALUES (1,'a');
+SELECT CONCAT(b,IF(a,'b','b')) FROM t1;
+CONCAT(b,IF(a,'b','b'))
+ab
+PREPARE stmt FROM "SELECT CONCAT(b,IF(a,?,?)) FROM t1";
+SET @b='b';
+EXECUTE stmt USING @b,@b;
+CONCAT(b,IF(a,?,?))
+ab
+SET @b='';
+EXECUTE stmt USING @b,@b;
+CONCAT(b,IF(a,?,?))
+a
+SET @b='я';
+EXECUTE stmt USING @b,@b;
+ERROR HY000: Illegal mix of collations (latin1_swedish_ci,IMPLICIT) and (utf8_general_ci,COERCIBLE) for operation 'concat'
+DEALLOCATE PREPARE stmt;
+DROP TABLE t1;
+#
 # End of 10.0 tests
 #

mysql-test/t/ctype_ucs.test

@@ -902,6 +902,13 @@ DROP TABLE t1;
 --echo #
 SELECT CONCAT(CONVERT('pi=' USING ucs2),PI()) AS PI;
 
+--echo #
+--echo # MDEV-6695 Bad column name for UCS2 string literals
+--echo #
+SET NAMES utf8, character_set_connection=ucs2;
+SELECT 'a','aa';
+
+
 --echo #
 --echo # End of 10.0 tests
 --echo #

mysql-test/t/ctype_utf8.test

@@ -1719,6 +1719,24 @@ SELECT CONCAT(a, IF(b>10, _utf8 X'61', _utf8 X'61')) FROM t1;
 SELECT CONCAT(a, IF(b>10, _utf8 X'61', _utf8 B'01100001')) FROM t1;
 DROP TABLE t1;
 
+--echo #
+--echo #  MDEV-6694 Illegal mix of collation with a PS parameter
+--echo #
+SET NAMES utf8;
+CREATE TABLE t1 (a INT, b VARCHAR(10) CHARACTER SET latin1);
+INSERT INTO t1 VALUES (1,'a');
+SELECT CONCAT(b,IF(a,'b','b')) FROM t1;
+PREPARE stmt FROM "SELECT CONCAT(b,IF(a,?,?)) FROM t1";
+SET @b='b';
+EXECUTE stmt USING @b,@b;
+SET @b='';
+EXECUTE stmt USING @b,@b;
+SET @b='я';
+--error ER_CANT_AGGREGATE_2COLLATIONS
+EXECUTE stmt USING @b,@b;
+DEALLOCATE PREPARE stmt;
+DROP TABLE t1;
+
 
 --echo #
 --echo # End of 10.0 tests

sql/item.cc

@@ -1073,10 +1073,14 @@ void Item::set_name(const char *str, uint length, CHARSET_INFO *cs)
     name_length= 0;
     return;
   }
-  if (cs->ctype)
-  {
-    const char *str_start= str;
 
+  const char *str_start= str;
+  if (!cs->ctype || cs->mbminlen > 1)
+  {
+    str+= cs->cset->scan(cs, str, str + length, MY_SEQ_SPACES);
+  }
+  else
+  {
     /*
       This will probably need a better implementation in the future:
       a function in CHARSET_INFO structure.
@@ -1086,21 +1090,21 @@ void Item::set_name(const char *str, uint length, CHARSET_INFO *cs)
       length--;
       str++;
     }
-    if (str != str_start && !is_autogenerated_name)
-    {
-      char buff[SAFE_NAME_LEN];
-      strmake(buff, str_start,
-              MY_MIN(sizeof(buff)-1, length + (int) (str-str_start)));
-
-      if (length == 0)
-        push_warning_printf(current_thd, Sql_condition::WARN_LEVEL_WARN,
-                            ER_NAME_BECOMES_EMPTY, ER(ER_NAME_BECOMES_EMPTY),
-                            buff);
-      else
-        push_warning_printf(current_thd, Sql_condition::WARN_LEVEL_WARN,
-                            ER_REMOVED_SPACES, ER(ER_REMOVED_SPACES),
-                            buff);
-    }
+  }
+  if (str != str_start && !is_autogenerated_name)
+  {
+    char buff[SAFE_NAME_LEN];
+    strmake(buff, str_start,
+            MY_MIN(sizeof(buff)-1, length + (int) (str-str_start)));
+
+    if (length == 0)
+      push_warning_printf(current_thd, Sql_condition::WARN_LEVEL_WARN,
+                          ER_NAME_BECOMES_EMPTY, ER(ER_NAME_BECOMES_EMPTY),
+                          buff);
+    else
+      push_warning_printf(current_thd, Sql_condition::WARN_LEVEL_WARN,
+                          ER_REMOVED_SPACES, ER(ER_REMOVED_SPACES),
+                          buff);
   }
   if (!my_charset_same(cs, system_charset_info))
   {
@@ -1269,27 +1273,11 @@ Item *Item_param::safe_charset_converter(CHARSET_INFO *tocs)
     SET @@arg= 1;
     EXECUTE stms USING @arg;
 
-    result_type is STRING_RESULT at prepare time,
+    In the above example result_type is STRING_RESULT at prepare time,
     and INT_RESULT at execution time.
   */
-  if (const_item())
-  {
-    if (state == NULL_VALUE)
-      return this;
-    uint cnv_errors;
-    String *ostr= val_str(&cnvstr);
-    if (!needs_charset_converter(tocs))
-      return this;
-    cnvitem->copy_value(ostr->ptr(), ostr->length(),
-                        ostr->charset(), tocs, &cnv_errors);
-    if (cnv_errors)
-       return NULL;
-    if (ostr->charset() == &my_charset_bin && tocs != &my_charset_bin &&
-        !cnvitem->check_well_formed_result(true))
-      return NULL;
-    return cnvitem;
-  }
-  return this;
+  return !const_item() || state == NULL_VALUE ?
+         this : const_charset_converter(tocs, true);
 }
 
 
@@ -3175,8 +3163,6 @@ Item_param::Item_param(uint pos_in_query_arg) :
     value is set.
   */
   maybe_null= 1;
-  cnvitem= new Item_string("", 0, &my_charset_bin, DERIVATION_COERCIBLE);
-  cnvstr.set(cnvbuf, sizeof(cnvbuf), &my_charset_bin);
 }
 
 
@@ -3736,18 +3722,14 @@ bool Item_param::convert_str_value(THD *thd)
       str_value.set_charset(value.cs_info.final_character_set_of_str_value);
     /* Here str_value is guaranteed to be in final_character_set_of_str_value */
 
-    max_length= str_value.numchars() * str_value.charset()->mbmaxlen;
-
-    /* For the strings converted to numeric form within some functions */
-    decimals= NOT_FIXED_DEC;
     /*
       str_value_ptr is returned from val_str(). It must be not alloced
       to prevent it's modification by val_str() invoker.
     */
     str_value_ptr.set(str_value.ptr(), str_value.length(),
                       str_value.charset());
-    /* Synchronize item charset with value charset */
-    collation.set(str_value.charset(), DERIVATION_COERCIBLE);
+    /* Synchronize item charset and length with value charset */
+    fix_charset_and_length_from_str_value(DERIVATION_COERCIBLE);
   }
   return rc;
 }
@@ -3777,7 +3759,8 @@ Item_param::clone_item()
   case STRING_VALUE:
   case LONG_DATA_VALUE:
     return new Item_string(name, str_value.c_ptr_quick(), str_value.length(),
-                           str_value.charset());
+                           str_value.charset(),
+                           collation.derivation, collation.repertoire);
   case TIME_VALUE:
     break;
   case NO_VALUE: