database/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-08-08 11:22:35 +03:00
Code Activity

5.1 version of a fix and test cases for bugs:

Browse Source 
Bug#4968 ""Stored procedure crash if cursor opened on altered table"
Bug#6895 "Prepared Statements: ALTER TABLE DROP COLUMN does nothing"
Bug#19182 "CREATE TABLE bar (m INT) SELECT n FROM foo; doesn't work from 
stored procedure."
Bug#19733 "Repeated alter, or repeated create/drop, fails"
Bug#22060 "ALTER TABLE x AUTO_INCREMENT=y in SP crashes server"
Bug#24879 "Prepared Statements: CREATE TABLE (UTF8 KEY) produces a 
growing key length" (this bug is not fixed in 5.0)

Re-execution of CREATE DATABASE, CREATE TABLE and ALTER TABLE 
statements in stored routines or as prepared statements caused
incorrect results (and crashes in versions prior to 5.0.25).

In 5.1 the problem occured only for CREATE DATABASE, CREATE TABLE
SELECT and CREATE TABLE with INDEX/DATA DIRECTOY options).
  
The problem of bugs 4968, 19733, 19282 and 6895 was that functions
mysql_prepare_table, mysql_create_table and mysql_alter_table are not
re-execution friendly: during their operation they modify contents
of LEX (members create_info, alter_info, key_list, create_list),
thus making the LEX unusable for the next execution.
In particular, these functions removed processed columns and keys from
create_list, key_list and drop_list. Search the code in sql_table.cc 
for drop_it.remove() and similar patterns to find evidence.
  
The fix is to supply to these functions a usable copy of each of the
above structures at every re-execution of an SQL statement. 
  
To simplify memory management, LEX::key_list and LEX::create_list
were added to LEX::alter_info, a fresh copy of which is created for
every execution.
  
The problem of crashing bug 22060 stemmed from the fact that the above 
metnioned functions were not only modifying HA_CREATE_INFO structure 
in LEX, but also were changing it to point to areas in volatile memory
of the execution memory root.
   
The patch solves this problem by creating and using an on-stack
copy of HA_CREATE_INFO in mysql_execute_command.

Additionally, this patch splits the part of mysql_alter_table
that analizes and rewrites information from the parser into
a separate function - mysql_prepare_alter_table, in analogy with
mysql_prepare_table, which is renamed to mysql_prepare_create_table.


mysql-test/r/ps.result:
  Update test results (Bug#19182, Bug#22060, Bug#4968, Bug#6895)
mysql-test/r/sp.result:
  Update results (Bug#19733)
mysql-test/t/ps.test:
  Add test cases for Bug#19182, Bug#22060, Bug#4968, Bug#6895
mysql-test/t/sp.test:
  Add a test case for Bug#19733
sql/field.h:
  Implement a deep copy constructor for create_field
sql/mysql_priv.h:
  LEX::key_list and LEX::create_list were moved to LEX::alter_info.
  Update declarations to use LEX::alter_info instead of these two
  members.
  Remove declarations of mysql_add_index, mysql_drop_index.
sql/sql_class.cc:
  Implement deep copy constructors.
sql/sql_class.h:
  Implement (almost) deep copy constructors for key_part_spec, 
  Alter_drop, Alter_column, Key, foreign_key.
  Replace pair<columns, keys> with an instance of Alter_info in
  select_create constructor. We create a new copy of Alter_info
  each time we re-execute SELECT .. CREATE prepared statement.
sql/sql_insert.cc:
  Adjust to a new signature of create_table_from_items.
sql/sql_lex.cc:
  Implement Alter_info::Alter_info that would make a "deep" copy
  of all definition lists (keys, columns).
  Move is_partition_management() from sql_partition.cc (feature-based
  file division is evil).
sql/sql_lex.h:
  Move key_list and create_list to class Alter_info. Implement
  Alter_info::Alter_info that can be used with PS and SP.
  Get rid of Alter_info::clear() which was an attempt to save on
  matches and always use Alter_info::reset().
  Implement an auxiliary Alter_info::init_for_create_from_alter()
  which is used in mysql_alter_table.
sql/sql_list.cc:
    Implement a copy constructor of class List that makes a deep copy
    of all list nodes.
sql/sql_list.h:
  Implement a way to make a deep copy of all list nodes.
sql/sql_parse.cc:
  Adjust to new signatures of mysql_create_table, mysql_alter_table,
  select_create. Functions mysql_create_index and mysql_drop_index has
  become identical after initialization of alter_info was moved to the 
  parser, and were merged. Flag enable_slow_log was not updated for 
  SQLCOM_DROP_INDEX, which was a bug.
  Just like CREATE INDEX, DROP INDEX is currently done via complete 
  table rebuild and is rightfully a slow administrative statement.
sql/sql_partition.cc:
  Move is_partition_management() to sql_lex.cc
  Adjust code to the new Alter_info.
sql/sql_table.cc:
  Adjust mysql_alter_table, mysql_recreate_table, mysql_create_table,
  mysql_prepare_table to new signatures.
  Rename mysql_prepare_table to mysql_prepare_create_table. Make
  sure it follows the convention and returns FALSE for success and
  TRUE for error.
  Move parts of mysql_alter_table to mysql_prepare_alter_table.
  Move the first invokation of mysql_prepare_table from mysql_alter_table
  to compare_tables, as it was needed only for the purpose
  of correct comparison.
  Since now Alter_info itself is created in the runtime mem root,
  adjust mysql_prepare_table to always allocate memory in the
  runtime memory root.
  Remove dead code.
sql/sql_yacc.yy:
  LEX::key_list and LEX::create_list moved to class Alter_info
This commit is contained in:
unknown
2007-05-28 15:30:01 +04:00
parent 850eadadc6
commit b11f1d0c97
17 changed files with 1602 additions and 1064 deletions
Download Patch File Download Diff File Expand all files Collapse all files

79
sql/sql_list.h
View File

@@ -1,3 +1,5 @@
#ifndef INCLUDES_MYSQL_SQL_LIST_H
#define INCLUDES_MYSQL_SQL_LIST_H
/* Copyright (C) 2000-2003 MySQL AB
This program is free software; you can redistribute it and/or modify
@@ -63,21 +65,24 @@ public:
pointer.
*/
class list_node :public Sql_alloc
/**
list_node - a node of a single-linked list.
@note We never call a destructor for instances of this class.
*/
struct list_node :public Sql_alloc
{
public:
list_node *next;
void *info;
list_node(void *info_par,list_node *next_par)
:next(next_par),info(info_par)
{}
{}
list_node() /* For end_of_list */
{
info=0;
next= this;
}
friend class base_list;
friend class base_list_iterator;
{
info= 0;
next= this;
}
};
@@ -93,12 +98,28 @@ public:
inline void empty() { elements=0; first= &end_of_list; last=&first;}
inline base_list() { empty(); }
/**
This is a shallow copy constructor that implicitly passes the ownership
from the source list to the new instance. The old instance is not
updated, so both objects end up sharing the same nodes. If one of
the instances then adds or removes a node, the other becomes out of
sync ('last' pointer), while still operational. Some old code uses and
relies on this behaviour. This logic is quite tricky: please do not use
it in any new code.
*/
inline base_list(const base_list &tmp) :Sql_alloc()
{
elements= tmp.elements;
first= tmp.first;
last= elements ? tmp.last : &first;
}
/**
Construct a deep copy of the argument in memory root mem_root.
The elements themselves are copied by pointer. If you also
need to copy elements by value, you should employ
list_copy_and_replace_each_value after creating a copy.
*/
base_list(const base_list &rhs, MEM_ROOT *mem_root);
inline base_list(bool error) { }
inline bool push_back(void *info)
{
@@ -185,6 +206,15 @@ public:
elements+= list->elements;
}
}
/**
Swap two lists.
*/
inline void swap(base_list &rhs)
{
swap_variables(list_node *, first, rhs.first);
swap_variables(list_node **, last, rhs.last);
swap_variables(uint, elements, rhs.elements);
}
inline list_node* last_node() { return *last; }
inline list_node* first_node() { return first;}
inline void *head() { return first->info; }
@@ -349,6 +379,8 @@ template <class T> class List :public base_list
public:
inline List() :base_list() {}
inline List(const List<T> &tmp) :base_list(tmp) {}
inline List(const List<T> &tmp, MEM_ROOT *mem_root) :
base_list(tmp, mem_root) {}
inline bool push_back(T *a) { return base_list::push_back(a); }
inline bool push_back(T *a, MEM_ROOT *mem_root)
{ return base_list::push_back(a, mem_root); }
@@ -547,3 +579,32 @@ public:
I_List_iterator(I_List<T> &a) : base_ilist_iterator(a) {}
inline T* operator++(int) { return (T*) base_ilist_iterator::next(); }
};
/**
Make a deep copy of each list element.
@note A template function and not a template method of class List
is employed because of explicit template instantiation:
in server code there are explicit instantiations of List<T> and
an explicit instantiation of a template requires that any method
of the instantiated class used in the template can be resolved.
Evidently not all template arguments have clone() method with
the right signature.
@return You must query the error state in THD for out-of-memory
situation after calling this function.
*/
template <typename T>
inline
void
list_copy_and_replace_each_value(List<T> &list, MEM_ROOT *mem_root)
{
/* Make a deep copy of each element */
List_iterator<T> it(list);
T *el;
while ((el= it++))
it.replace(el->clone(mem_root));
}
#endif // INCLUDES_MYSQL_SQL_LIST_H