[pylucene-dev] Fwd: RE: GC Warning: Header allocation failed: Dropping block.

Andi Vajda vajda at osafoundation.org
Thu Feb 17 13:05:14 PST 2005


> Well, I can try to experiment myself if you can't reproduce bix index
> on your computer...
>
> But I do not know what should I try to tweak about gcj and GC :)
> Tell me if you have ideas

You and I know exactly the same things now. I do not have a big index 
available although a PyLucene user sent me a DVD with lots of data to create 
one. I'm about to take off for a week of vacation and have to get ready for 
PyCON when I return so I don't think I'll be able to spend any reasonable 
amount of time on this issue until April.

> Maybe we need to ask Boehm, Hans?

I think he is the person who can help the most effectively with this issue.

Andi..

>
> AV> Andi..
>
>>>
>>> So, it seems like pure Java works fine.
>>>
>>> 2. Also I have tried to run PyLucene source with environment
>>> GC_PRINT_STATS=1 as Boehm, Hans has suggested.
>>> I received following log of messages (gc.log). Please, check it. Maybe
>>> it will give you ideas...
>>>
>>> 3. Also I have tried 'export GC_FIND_LEAK=1'. Program shown some
>>> messages. I hope this will give us ideas.
>>>
>>> Leaked composite object at start: 0x82f69d0, appr. length: 16
>>> ...
>>> Leaked composite object at start: 0x82f6d60, appr. length: 16
>>> Leaked atomic object at start: 0x82f5f50, appr. length: 112
>>> Leaked composite object at start: 0x82f4fb0, appr. length: 16
>>> Leaked composite object at start: 0x82f4fc0, appr. length: 16
>>> Leaked composite object at start: 0x82f4ff0, appr. length: 16
>>> Leaked atomic object at start: 0x82f1500, appr. length: 80
>>> ...
>>> Leaked composite object at start: 0x82c12d0, appr. length: 80
>>> Leaked atomic object at start: 0x8293fa8, appr. length: 8
>>> ...
>>> Leaked atomic object at start: 0x8293fd8, appr. length: 8
>>> Leaked composite object at start: 0x826c930, appr. length: 24
>>> ...
>>> Leaked composite object at start: 0x82462d0, appr. length: 24
>>>
>>> Yura Smolsky.
>>>
>
>
>
>
> Yura Smolsky.
>
>
>


More information about the pylucene-dev mailing list