Bug #62111 [Opn->Ver]: MySQL PDO memory leaks, when used own result row class

From: Date: Tue, 10 Mar 2020 15:26:07 +0000
Subject: Bug #62111 [Opn->Ver]: MySQL PDO memory leaks, when used own result row class
References: 1  Groups: php.bugs 
Request: Send a blank email to php-bugs+get-226010@lists.php.net to get a copy of this message
Edit report at https://bugs.php.net/bug.php?id=62111&edit=1

 ID:                 62111
 Updated by:         cmb@php.net
 Reported by:        hosiplan at gmail dot com
 Summary:            MySQL PDO memory leaks, when used own result row
                     class
-Status:             Open
+Status:             Verified
 Type:               Bug
 Package:            PDO MySQL
 Operating System:   Linux
 PHP Version:        5.4.4RC1
 Block user comment: N
 Private report:     N

 New Comment:

I can confirm that we're still leaking the pdo_stmt_t objects:

==32703== 3,680 bytes in 10 blocks are definitely lost in loss record 1 of 1
==32703==    at 0x4C2BBAF: malloc (vg_replace_malloc.c:299)
==32703==    by 0x4D52B6: __zend_malloc (zend_alloc.c:2976)
==32703==    by 0x4D423A: _malloc_custom (zend_alloc.c:2417)
==32703==    by 0x4D4368: _emalloc (zend_alloc.c:2536)
==32703==    by 0x3291C1: zend_object_alloc (zend_objects_API.h:91)
==32703==    by 0x331151: pdo_dbstmt_new (pdo_stmt.c:2314)
==32703==    by 0x511762: _object_and_properties_init (zend_API.c:1417)
==32703==    by 0x5117D3: object_init_ex (zend_API.c:1431)
==32703==    by 0x323948: pdo_stmt_instantiate (pdo_dbh.c:412)
==32703==    by 0x324290: zim_PDO_prepare (pdo_dbh.c:522)
==32703==    by 0x5781A9: ZEND_DO_FCALL_SPEC_RETVAL_USED_HANDLER (zend_vm_execute.h:1730)
==32703==    by 0x5D6FF6: execute_ex (zend_vm_execute.h:53821)


Previous Comments:
------------------------------------------------------------------------
[2012-12-15 20:35:10] enumag at gmail dot com

Try this one instead. It does not matter whether you use MySQL or SQLlite. The problem is that if
you add some arguments for the constructor, it consumes more and more memory without any reason.

---------
class pdo_row {
	public function __construct() { }
}

$pdo = new PDO('mysql:host=127.0.0.1;dbname=information_schema', 'root',
'');

$pdo->exec("DROP TABLE test");
$pdo->exec("CREATE TABLE test(id INT, col VARCHAR(200))");
for ($i = 0; $i < 100; $i++) {
  $pdo->exec(sprintf("INSERT INTO test(id, col) VALUES (1,
'012345678901234567890123456789012345678901234567890123456789-%d')", $i));
}

printf("With ctor argument (memory usage increase):");
for ($i = 0; $i < 10; $i++) {
	$stmt = $pdo->prepare("SELECT col FROM test");
    $stmt->setFetchMode(PDO::FETCH_CLASS, 'pdo_row', array($stmt));
	$stmt->execute();
	$rows = $stmt->fetchAll();
	printf("emalloc %d kB, malloc %d kB\n",
		memory_get_usage() / 1024,
		memory_get_usage(true) / 1024);
}

printf("Without ctor argument (no memory usage increase):");
for ($i = 0; $i < 10; $i++) {
	$stmt = $pdo->prepare("SELECT col FROM test");
    $stmt->setFetchMode(PDO::FETCH_CLASS, 'pdo_row');
	$stmt->execute();
	$rows = $stmt->fetchAll();
	printf("emalloc %d kB, malloc %d kB\n",
		memory_get_usage() / 1024,
		memory_get_usage(true) / 1024);
}

------------------------------------------------------------------------
[2012-06-13 15:10:15] uw@php.net

It matters a whole lot what the cause is. It is not a leak. First, it helps to have some clean test
code, such as this:

class pdo_row {
	public function __construct($stmt) {
		$stmt = NULL;
	}
}

/* $pdo = new PDO("mysql:dbname=test;unix_socket=/var/run/mysql/mysql.sock",
"root", ""); */
$pdo = new PDO("sqlite::memory:");

$pdo->exec("DROP TABLE test");
$pdo->exec("CREATE TABLE test(id INT, col VARCHAR(200))");
for ($i = 0; $i < 100; $i++) {
  $pdo->exec(sprintf("INSERT INTO test(id, col) VALUES (1,
'012345678901234567890123456789012345678901234567890123456789-%d')", $i));
}

for ($i = 0; $i < 10; $i++) {
	$stmt = $pdo->prepare("SELECT col FROM test");
	$stmt->execute();
	$stmt->setFetchMode(PDO::FETCH_CLASS, 'pdo_row', array($stmt));
	$rows = $stmt->fetchAll();
	printf("emalloc %d kB, malloc %d kB\n",
		memory_get_usage() / 1024,
		memory_get_usage(true) / 1024);
}


Running the above for SQLlite gives me:

emalloc 205 kB, malloc 256 kB
emalloc 206 kB, malloc 512 kB
emalloc 207 kB, malloc 512 kB
emalloc 208 kB, malloc 512 kB
emalloc 209 kB, malloc 512 kB
emalloc 210 kB, malloc 512 kB
emalloc 211 kB, malloc 512 kB
emalloc 212 kB, malloc 512 kB
emalloc 213 kB, malloc 512 kB
emalloc 214 kB, malloc 512 kB


Running the above for MySQL gives me:



emalloc 221 kB, malloc 256 kB
emalloc 231 kB, malloc 512 kB
emalloc 240 kB, malloc 512 kB
emalloc 250 kB, malloc 512 kB
emalloc 259 kB, malloc 512 kB
emalloc 269 kB, malloc 512 kB
emalloc 278 kB, malloc 512 kB
emalloc 288 kB, malloc 512 kB
emalloc 297 kB, malloc 512 kB
emalloc 307 kB, malloc 512 kB


Second, it helps to understand the matter. In both cases emalloc() figures increase, which is to be
expected. PHP objects are created, the extensions allocate memory using the PHP internal e*alloc()
function family and the figures increase. And, in both cases malloc() figures are the same. 

No difference between MySQL and SQLite.

------------------------------------------------------------------------
[2012-06-12 16:18:47] hosiplan at gmail dot com

I don't really care what you name it. It's a real problem and its eating my 
memory and it shouldn't!

------------------------------------------------------------------------
[2012-06-12 13:33:16] uw@php.net

There is no leak with MySQL. Memory usage increases, that's it.

==6216== LEAK SUMMARY:
==6216==    definitely lost: 0 bytes in 0 blocks
==6216==    indirectly lost: 0 bytes in 0 blocks
==6216==      possibly lost: 0 bytes in 0 blocks
==6216==    still reachable: 54 bytes in 2 blocks
==6216==         suppressed: 0 bytes in 0 blocks
==6216== Rerun with --leak-check=full to see details of leaked memory

------------------------------------------------------------------------
[2012-06-11 13:48:11] juzna dot cz at gmail dot com

The same causes PHP to crash completely on Windows 32bit with MSSQL server (using 
sqlsrv driver). I guess the root cause will be related.

------------------------------------------------------------------------


The remainder of the comments for this report are too long. To view
the rest of the comments, please view the bug report online at

    https://bugs.php.net/bug.php?id=62111


--
Edit this bug report at https://bugs.php.net/bug.php?id=62111&edit=1


Thread (11 messages)

« previous php.bugs (#226010) next »