Everything

php

Solved bug #24842, memory corruption

php.internals

Vesselin Atanasov

23 years ago

Hello.
I found the cause for bug #24842. In function shutdown_executor(), the
"arg_types_stack" stack is
cleaned way too early, and later when some destructor calls a function like
printf(), the helper function
zend_do_fcall_handler() uses the "arg_types_stack", which is freed earlier,
thus overwriting any data
that has replaced the arg_types_stack->elements memory block. A small patch
follows:

diff -ruN php5-200307300330.orig/Zend/zend_execute_API.c
php5-200307300330/Zend/zend_execute_API.c
--- php5-200307300330.orig/Zend/zend_execute_API.c	2003-07-27
17:07:14.000000000 +0000
+++ php5-200307300330/Zend/zend_execute_API.c	2003-07-30 07:55:39.000000000
+0000
@@ -189,7 +189,10 @@
void shutdown_executor(TSRMLS_D)
{
zend_try {
+/* Moved after symbol table cleaners because arg_types_stack is used by
zend_do_fcall_handler(), so if a
+   destructor calls a function like printf() it will cause memory
corruption
zend_ptr_stack_destroy(&EG(arg_types_stack));
+ */

/* Removed because this can not be safely done, e.g. in this situation:
Object 1 creates object 2
@@ -286,6 +289,7 @@

zend_hash_destroy(&EG(included_files));

+		zend_ptr_stack_destroy(&EG(arg_types_stack));
zend_ptr_stack_destroy(&EG(user_error_handlers));
zend_ptr_stack_destroy(&EG(user_exception_handlers));
zend_objects_store_destroy(&EG(objects_store));

John Coggeshall

23 years ago

Hey all:

I've written a new extension for PHP (ZE2 only) Based on the famed HTML
Tidy's (http://tidy.sf.net/) library. This extension provides more than
just an incredibly easy way to clean and repair HTML documents, and
includes API for traversing an arbitrary HTML document using the ZE2 OO
support.

I've put the extension, basic PHPDocs, working examples, tests, and of
course the extension itself on my web site:

http://www.coggeshall.org/php/php-tidy-0-5b.tar.gz

Although I haven't written true PHPDocs yet, README_TIDY in the tidy/
directory outlines the API including a description of the OO methods and
proprieties available to accessing the parsed HTML document tree.

I am interested in hearing from the internal@ community on my extension
and finding out what everyone thinks of it. There are memleaks which
still need to be tracked down (which I believe either have to do with
ZE2 itself or because I am missing something in my OO implementation),
and I know there are probably bugs.. I'd welcome suggestions, patches,
and even just "it broke doing this".

I plan on maintaining this extension on my web site and perhaps PECL
(unless of course this is something worthy for the standard PHP5
distro), so if nothing else you can always find it there.

Regards,

John

PS -- Here is a paste of one of the examples for those who are curious
to how the OO stuff works (pulls all <A HREF> links out):

<?php

/* Create a Tidy Resource */
$tidy = tidy_create();

/* Parse the document */
tidy_parse_file($tidy, $_SERVER['argv'][1]);

/* Fix up the document */
tidy_clean_repair($tidy);

/* Get an object representing everything from the <HTML> tag in */
$html = tidy_get_html($tidy);

/* Traverse the document tree */
print_r(get_links($html));

function get_links($node) {
$urls = array();

/* Check to see if we are on an <A> tag or not */
if($node->id == TIDY_TAG_A) {
/* If we are, find the HREF attribute */
$attrib = $node->get_attr_type(TIDY_ATTR_HREF);
if($attrib) {
/* Add the value of the HREF attrib to $urls */
$urls[] = $attrib->value;
}

}

/* Are there any children? */
if($node->has_children()) {

/* Traverse down each child recursively */
foreach($node->children as $child) {

/* Append the results from recursion to $urls */
foreach(get_links($child) as $url) {

$urls[] = $url;

}

}
}

return $urls;
}

?>

--
-~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~-
John Coggeshall
john at coggeshall dot org                 http://www.coggeshall.org/
-~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~-

(Marcus Börger)

23 years ago

Hello John,

Wednesday, July 30, 2003, 1:26:02 PM, you wrote:

JC> Hey all:

JC> I've written a new extension for PHP (ZE2 only) Based on the famed HTML
JC> Tidy's (http://tidy.sf.net/) library. This extension provides more than
JC> just an incredibly easy way to clean and repair HTML documents, and
JC> includes API for traversing an arbitrary HTML document using the ZE2 OO
JC> support.

JC> I've put the extension, basic PHPDocs, working examples, tests, and of
JC> course the extension itself on my web site:

JC> http://www.coggeshall.org/php/php-tidy-0-5b.tar.gz

JC> Although I haven't written true PHPDocs yet, README_TIDY in the tidy/
JC> directory outlines the API including a description of the OO methods and
JC> proprieties available to accessing the parsed HTML document tree.

JC> I am interested in hearing from the internal@ community on my extension
JC> and finding out what everyone thinks of it. There are memleaks which
JC> still need to be tracked down (which I believe either have to do with
JC> ZE2 itself or because I am missing something in my OO implementation),
JC> and I know there are probably bugs.. I'd welcome suggestions, patches,
JC> and even just "it broke doing this".

JC> I plan on maintaining this extension on my web site and perhaps PECL
JC> (unless of course this is something worthy for the standard PHP5
JC> distro), so if nothing else you can always find it there.

Some months ago i had a similar idea, mainly to have it as an output handler.
So in general i think it's a cool idea. So my question is will you provide
such an output handler?

Besides that i think you should put it into PECL :-)

--
Best regards,
Marcus                            mailto:helly@php.net

Unnamed Person

23 years ago

"Marcus BöRger" <marcus.boerger@t-online.de> a écrit dans le message de
news:992327984.20030730202356@post.rwth-aachen.de...

> Hello John,
>
> Wednesday, July 30, 2003, 1:26:02 PM, you wrote:
>
> JC> Hey all:
>
> JC> I've written a new extension for PHP (ZE2 only) Based on the famed

HTML

> JC> Tidy's (http://tidy.sf.net/) library. This extension provides more

than

> JC> just an incredibly easy way to clean and repair HTML documents, and
> JC> includes API for traversing an arbitrary HTML document using the ZE2

OO

> JC> support.
>
> JC> I've put the extension, basic PHPDocs, working examples, tests, and of
> JC> course the extension itself on my web site:
>
> JC> http://www.coggeshall.org/php/php-tidy-0-5b.tar.gz
>
> JC> Although I haven't written true PHPDocs yet, README_TIDY in the tidy/
> JC> directory outlines the API including a description of the OO methods

and

> JC> proprieties available to accessing the parsed HTML document tree.
>
> JC> I am interested in hearing from the internal@ community on my

extension

> JC> and finding out what everyone thinks of it. There are memleaks which
> JC> still need to be tracked down (which I believe either have to do with
> JC> ZE2 itself or because I am missing something in my OO implementation),
> JC> and I know there are probably bugs.. I'd welcome suggestions, patches,
> JC> and even just "it broke doing this".
>
> JC> I plan on maintaining this extension on my web site and perhaps PECL
> JC> (unless of course this is something worthy for the standard PHP5
> JC> distro), so if nothing else you can always find it there.
>
> Some months ago i had a similar idea, mainly to have it as an output

handler.

> So in general i think it's a cool idea. So my question is will you provide
> such an output handler?
>
> Besides that i think you should put it into PECL :-)

+1 too.

>
>
>
> --
> Best regards,
>  Marcus                            mailto:helly@php.net
>

--
Regards.
M.CHAILLAN Nicolas
nicos@php.net
www.WorldAKT.com Hébergement de sites internets.

John Coggeshall

23 years ago

> Some months ago i had a similar idea, mainly to have it as an output handler.
> So in general i think it's a cool idea. So my question is will you provide
> such an output handler?

I actually was bugging Wez about filters for this exact reason. :)

> Besides that i think you should put it into PECL :-)

just requested an account.

John

--
-~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~-
John Coggeshall
john at coggeshall dot org                 http://www.coggeshall.org/
-~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~--~=~-