Re: [Ming-dev] Memory problem with the php extension ?

[Alan B. <abo...@ma...>]

Yes Yoav, your email has been received.

Alan

Yoav Levy wrote:
> Hi,
>
>
> Yesterday I sent the message below; I am new to this newsletter, 
> please let me know if I sent it correctly, thanks a lot!
>
>
> I made a ming upgrade on my server. Whenever I create an action (new 
> SWFAction()) longer then 2000 characters I get :
> child pid 2608 exit signal Segmentation fault (11).
> I suspect a memory leak is causing this problem and it might be 
> related to the problem discussed here.
>
> Thanks,
> Yoav
>
>
>> Absolutly,
>>
>> It's really in the php extension. I'm wondering what happen.. if the  
>> object reference is properly removed for the list and the C 
>> destructor is well called, where could be the problem inside the 
>> extension? This could be inside php code. I don't think that 
>> extensions that use the new zend api have this problem.
>>
>> Alan
>>
>> -----Original Message-----
>> From: "Klaus Rechert" <kla...@rz...>
>> Sent: Wednesday, October 29, 2008 6:00am
>> To: "Alan Boudreault" <abo...@ma...>, "Ming Developers 
>> mailing list" <min...@li...>
>> Subject: Re: [Ming-dev] Memory problem with the php extension ?
>>
>> Hi,
>>
>> sorry I was mostly offline the last few days. I poked a little bit in 
>> the code but haven't found a real solution yet.
>>
>> For sure is:
>>     1. ming's shape() does not leak (checked via a C testcase)
>>     2. memory usage scales with the number of instances
>>
>> I have the strong suspicion that the list holding the instances is 
>> not cleared.
>>
>> Cheers
>>     Klaus
>>
>>  
>>> Hi Klaus,
>>>
>>> I'm wondering if you have any new about the problem ? (Can you add 
>>> me in CC in the bugzilla plz)
>>>
>>> Thanks,
>>> Alan
>>>
>>> Alan Boudreault wrote:
>>>    
>>>> In fact, replacing the zend_list_insert by 
>>>> ZEND_REGISTER_RESOURCE........ has the same effect that removing 
>>>> the zend_list_addref().
>>>>
>>>> Alan
>>>>
>>>> Klaus Rechert wrote:
>>>>      
>>>>> Hi Alan,
>>>>>
>>>>> the __construct function looks like this now:
>>>>>
>>>>> PHP_METHOD(swfshape, __construct)
>>>>> {
>>>>>    SWFShape shape = newSWFShape();
>>>>>    int rsrc_id = ZEND_REGISTER_RESOURCE(return_value, shape, 
>>>>> le_swfshapep);
>>>>> //    int rsrc_id = zend_list_insert(shape, le_swfshapep);
>>>>>    object_init_ex(getThis(), shape_class_entry_ptr);
>>>>>    add_property_resource(getThis(), "shape", rsrc_id);
>>>>>    zend_list_addref(rsrc_id);
>>>>> }
>>>>>
>>>>> Commenting the zend_list_addref() has no additional effect on 
>>>>> memory usage here.
>>>>>
>>>>> Thanks
>>>>>    Klaus
>>>>>
>>>>>        
>>>>>> Hi Klaus,
>>>>>>
>>>>>> I was writing you a comment on bugzilla.... but still waiting the 
>>>>>> account creation confirmation... here's what i wrote:
>>>>>>
>>>>>> If you comment the following line, in PHP_METHOD(swfshape, 
>>>>>> __construct) method:
>>>>>> zend_list_addref(ret);
>>>>>>
>>>>>> The destructor is well called (verified with a printf in the 
>>>>>> destructor). BUT, the PHP OBJECT doesn't seem to be freed by 
>>>>>> php   garbage collector. See what i get now:
>>>>>>
>>>>>> Beginning of script : 57368 bytes
>>>>>>
>>>>>> End of script : 769960 bytes
>>>>>>
>>>>>> So, there are still "memory leaks".
>>>>>>
>>>>>> Thanks,
>>>>>> Alan
>>>>>>
>>>>>> Klaus Rechert wrote:
>>>>>>  
>>>>>>          
>>>>>>> It is me again ;)
>>>>>>>
>>>>>>> Just dug a little bit in ZEND sources. Using 
>>>>>>> ZEND_REGISTER_RESOURCE() instead of zend_list_insert() made the 
>>>>>>> destructor function working. Running again your testcase results 
>>>>>>> in:
>>>>>>>
>>>>>>> End of script : 1012316 bytes
>>>>>>>
>>>>>>> instead of
>>>>>>>
>>>>>>> End of script : 1524912 bytes for Ming 0.4.2.
>>>>>>>
>>>>>>> One little step... :)
>>>>>>>
>>>>>>> Cheers
>>>>>>>    Klaus
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>  
>>>>>>>            
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> just checked your example. You are right PHP leaks. I'm not 
>>>>>>>> sure why... There are destructor functions defined and 
>>>>>>>> registered, but not called at all.
>>>>>>>>
>>>>>>>> I filed a bug
>>>>>>>> http://bugs.libming.net/show_bug.cgi?id=74
>>>>>>>>
>>>>>>>> Thanks
>>>>>>>>     Klaus
>>>>>>>>
>>>>>>>>
>>>>>>>>  
>>>>>>>>                 
>>>>>>>>> Hi guys, i'm wondering if you are aware of this problem using 
>>>>>>>>> your extension (or in the php garbage collector) ?
>>>>>>>>>
>>>>>>>>> See my script and results: http://pastebin.ca/1231918
>>>>>>>>>
>>>>>>>>> Do you know why the memory of the php object are not freed ?
>>>>>>>>>
>>>>>>>>> Thanks,
>>>>>>>>> Alan
>>>>>>>>>
>>>>>>>>>                               
>>>>>>>> ------------------------------------------------------------------------- 
>>>>>>>>
>>>>>>>> This SF.Net email is sponsored by the Moblin Your Move 
>>>>>>>> Developer's challenge
>>>>>>>> Build the coolest Linux based applications with Moblin SDK & 
>>>>>>>> win great prizes
>>>>>>>> Grand prize is a trip for two to an Open Source event anywhere 
>>>>>>>> in the world
>>>>>>>> http://moblin-contest.org/redirect.php?banner_id=100&url=/
>>>>>>>> _______________________________________________
>>>>>>>> Ming-devr mailing list
>>>>>>>> Min...@li...
>>>>>>>> https://lists.sourceforge.net/lists/listinfo/ming-devr
>>>>>>>>                       
>>>>>>             
>>>>       
>>>     
>>
>>
>>
>>
>> ------------------------------------------------------------------------- 
>>
>> This SF.Net email is sponsored by the Moblin Your Move Developer's 
>> challenge
>> Build the coolest Linux based applications with Moblin SDK & win 
>> great prizes
>> Grand prize is a trip for two to an Open Source event anywhere in the 
>> world
>> http://moblin-contest.org/redirect.php?banner_id=100&url=/
>> _______________________________________________
>> Ming-devr mailing list
>> Min...@li...
>> https://lists.sourceforge.net/lists/listinfo/ming-devr
>>
>>   
>


-- 
Alan Boudreault
Mapgears
http://www.mapgears.com