commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Brett Porter <br...@apache.org>
Subject Re: [jelly] Maven JSL memory leak and Jelly
Date Sun, 30 Jan 2005 02:13:37 GMT
I tried out the new patch as provided by dIon on the list, and it 
completely toasted Maven. Sorry.

I don't have much time to investigate this presently - it will show up 
during bootstrap if you want to try that, though the error is cryptic 
(goals are not found, so its really affecting the deep internals of 
Maven's Jelly/Werkz integration).

- Brett

Hans Gilde wrote:

>FYI, this is definitely a different problem. It very well could be solved by
>my new patch, if not I'll check it out.
>
>-----Original Message-----
>From: Brett Porter [mailto:brett@apache.org] 
>Sent: Thursday, January 27, 2005 6:36 AM
>To: Jakarta Commons Developers List
>Subject: Re: [jelly] Maven JSL memory leak and Jelly
>
>The good news: nailed it in xdoc. If you run multiproject:goal 
>-Dgoal=xdoc, you only need about 13mb of memory for any number of projects.
>
>Tha bad news: multiproject:goal -Dgoal=site still leaks another 4-6mb 
>per project with the default set of reports. With no reports, none is 
>leaked, but it is no individual report that is causing it - just a 
>cumulative effect.
>
>As all that is processed in Jelly, there's a strong possiblity that this 
>is in Jelly somewhere else again, however I've not got any conclusive 
>proof at this point.
>
>I'll try your new patch and let you know if it makes any difference, and 
>possibly do some more investigations later after I've got dom4j/jaxen 
>working on Maven's trunk again.
>
>I'd say this is "good enough" for RC2, as it has at least halved the 
>consumption of building a site. Nice work!
>
>- Brett
>
>Hans Gilde wrote:
>
>  
>
>>It was just a little issue, it should be fixed now.
>>
>>-----Original Message-----
>>From: Brett Porter [mailto:brett@apache.org] 
>>Sent: Monday, January 24, 2005 10:09 PM
>>To: Jakarta Commons Developers List
>>Subject: RE: [jelly] Maven JSL memory leak and Jelly
>>
>>Thanks Hans!
>>
>>That's what I was thinking of doing, so I'm glad you've got it under
>>control.
>>
>>When I was testing, the size of the included JSL affected the amount of
>>memory
>>leaked, so I guess it is tag caching again?
>>
>>I tried a context.clear[Thread]ScriptData after the execution of runScript
>>in
>>include, but that didn't help. Unfortunately that's all I had time for this
>>morning before work.
>>
>>Good luck...
>>
>>Cheers,
>>Brett
>>
>>Quoting Hans Gilde <hgilde-commons@earthlink.net>:
>>
>> 
>>
>>    
>>
>>>I just added a super-simple test case for the include tag to the core test
>>>suite. Include is a special kind of tag, so hopefully it's the only one
>>>leaking. Once we fix this problem, a retest will definitely be in order.
>>>
>>>-----Original Message-----
>>>From: Brett Porter [mailto:brett@apache.org] 
>>>Sent: Monday, January 24, 2005 9:47 PM
>>>To: Jakarta Commons Developers List
>>>Subject: RE: [jelly] Maven JSL memory leak and Jelly
>>>
>>>Yup. AFAICT, it's been there since it's creation.
>>>
>>>Do you need some additional info from me? I'm happy to try to make a
>>>   
>>>
>>>      
>>>
>>smaller
>> 
>>
>>    
>>
>>>test case and/or test with Maven later if required.
>>>
>>>- Brett
>>>
>>>Quoting Hans Gilde <hgilde-commons@earthlink.net>:
>>>
>>>   
>>>
>>>      
>>>
>>>>There's definitely a leak in the include tag. 
>>>>
>>>>Brett, I'm guessing that this leak also exists pre-RC2, right?
>>>>
>>>>-----Original Message-----
>>>>From: Brett Porter [mailto:brett@apache.org] 
>>>>Sent: Monday, January 24, 2005 7:30 AM
>>>>To: Jakarta Commons Developers List
>>>>Subject: Re: [jelly] Maven JSL memory leak and Jelly
>>>>
>>>>Paul Libbrecht wrote:
>>>>
>>>>     
>>>>
>>>>        
>>>>
>>>>>Ideally, a test-case would be awesome, even if it refers to some  
>>>>>far-away stylesheet...
>>>>>       
>>>>>
>>>>>          
>>>>>
>>>>I'll try and narrow it down first by cutting down site.jsl to the 
>>>>minimum that leaks.
>>>>
>>>>     
>>>>
>>>>        
>>>>
>>>>>Can you try calling .clear() on the result of this
>>>>>               context.runScript(uri, output, isExport(),
>>>>>       
>>>>>
>>>>>          
>>>>>
>>isInherit())
>> 
>>
>>    
>>
>>>>>(and the other call).
>>>>>       
>>>>>
>>>>>          
>>>>>
>>>>so try with context.clear() inside IncludeTag? Ok, will try tomorrow.
>>>>
>>>>     
>>>>
>>>>        
>>>>
>>>>>Maybe that'll help.
>>>>>In all cases, this context is gc-ed shortly after, I believe... so I 

>>>>>see no reasons for big leaks at the tag-cache level.
>>>>>       
>>>>>
>>>>>          
>>>>>
>>>>It leaks 200k every time site.jsl is run, so should be easy to find, 
>>>>right? :)
>>>>
>>>>     
>>>>
>>>>        
>>>>
>>>>>Also, maybe it would help to give more details where to go... I think
>>>>>       
>>>>>
>>>>>          
>>>>>
>> 
>>
>>    
>>
>>>>>this was reported about very long ago so maybe a distribution  
>>>>>maven-1.0.2 or such should have this bug ?
>>>>>       
>>>>>
>>>>>          
>>>>>
>>>>easiest to run it against Maven's trunk really as it is compatible with 
>>>>Jelly RC2 (RC1 was broken, so it currently uses beta-4 until RC2 is 
>>>>released). RC2 won't work with Maven 1.0.x.
>>>>
>>>>If you want to give it a go:
>>>>- check out Maven from SVN and build:
>>>>   svn co http://svn.apache.org/repos/asf/maven/maven-1/core/trunk 
>>>>maven-1/core/trunk
>>>>   svn co http://svn.apache.org/repos/asf/maven/maven-1/plugins/trunk 
>>>>maven-1/plugins/trunk
>>>>- build maven inside maven-1/core/trunk (see 
>>>>
>>>>     
>>>>
>>>>        
>>>>
>>http://www.apache.org/~brett/maven-stage-site/developers/building-from-sour
>>    
>>
>c
>  
>
>> 
>>
>>    
>>
>>>>e.html, 
>>>>make sure you reassign a new MAVEN_HOME!)
>>>>- run "maven xdoc" and watch memory go...
>>>>- twiddle xdoc plugin and run "maven plugin:install" to test. Use the 
>>>>memory tag lib I just added to jelly to test. <memory:gc/> and 
>>>><memory:showFree />.
>>>>
>>>>Hopefully I'll have more information tomorrow, I'm going to bed now. I'm
>>>>     
>>>>
>>>>        
>>>>
>> 
>>
>>    
>>
>>>>keen to kill this before Jelly 1.0 if possible.
>>>>
>>>>Thanks,
>>>>Brett
>>>>
>>>>
>>>>---------------------------------------------------------------------
>>>>To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
>>>>For additional commands, e-mail: commons-dev-help@jakarta.apache.org
>>>>
>>>>
>>>>---------------------------------------------------------------------
>>>>To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
>>>>For additional commands, e-mail: commons-dev-help@jakarta.apache.org
>>>>
>>>>     
>>>>
>>>>        
>>>>
>>>---------------------------------------------------------------------
>>>To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
>>>For additional commands, e-mail: commons-dev-help@jakarta.apache.org
>>>
>>>
>>>---------------------------------------------------------------------
>>>To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
>>>For additional commands, e-mail: commons-dev-help@jakarta.apache.org
>>>
>>>   
>>>
>>>      
>>>
>>
>>
>>---------------------------------------------------------------------
>>To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
>>For additional commands, e-mail: commons-dev-help@jakarta.apache.org
>>
>>
>>---------------------------------------------------------------------
>>To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
>>For additional commands, e-mail: commons-dev-help@jakarta.apache.org
>>
>>
>> 
>>
>>    
>>
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
>For additional commands, e-mail: commons-dev-help@jakarta.apache.org
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
>For additional commands, e-mail: commons-dev-help@jakarta.apache.org
>
>
>  
>


---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org


Mime
View raw message