Ferret talk - Oct 2006 - Memory allocation bug with index.search

Hi Dave ( again ! )

I''ve been searching for a while into my extension code to understand 
what was the matter, and Florent Solt who hasn''t got my extension has 
the same problem, so it figured to be a ferret problem.. Unfortunately, 
we''re unable at the moment to reproduce it in a little code so you
could
debug easilier... I''m working on it.

Heres the issue :

In the index, there are a lot of stuff with a type field, and we''re 
experiencing the same bug with just type => hardware or mixed datas. 
We''ve got around 12400 docs with this type, here are the queries (THEY 
ARE ALL LAUNCHED IN A _NEW_ INSTANCE OF FERRET ! (with the same index of 
course)):

this one works fine :
>> INDEX.search(''type:hardware'').to_s
=> "TopDocs: total_hits = 12490, max_score = 1.751220 [\n\t13997 
\"61426\": 1.751220\n\t13998 \"61427\": 1.751220\n\t13999
\"61428\":
1.751220\n\t14000 \"61429\": 1.751220\n\t14001 \"61430\": 
1.751220\n\t14002 \"61431\": 1.751220\n\t14003 \"61432\": 
1.751220\n\t14004 \"61433\": 1.751220\n\t14005 \"61434\": 
1.751220\n\t14006 \"61435\": 1.751220\n]\n"

and this one doesn''t.. :
>> INDEX.search(''type:hardware'', :limit => :all).to_s
/usr/local/lib/site_ruby/1.8/ferret/index.rb:718: [BUG] rb_gc_mark(): 
unknown data type 0x18(0x89b4268) non objectruby 1.8.4 (2005-12-24) 
[i486-linux]
Aborted

i tried to get to the limit point and got it :
works :
>> INDEX.search(''type:hardware'', :limit => 9719).to_s
don''t work :
>> INDEX.search(''type:hardware'', :limit => 9720).to_s
/usr/local/lib/site_ruby/1.8/ferret/index.rb:718: [BUG] Segmentation 
fault
ruby 1.8.4 (2005-12-24) [i486-linux]

BUT ! if, IN THE SAME INSTANCE OF RUBY, i do the first and then the 
second one, it works !
>> INDEX.search(''type:hardware'', :limit => 9719).to_s
...works
>> INDEX.search(''type:hardware'', :limit => 9720).to_s
...works
>> INDEX.search(''type:hardware'', :limit => :all).to_s
WORKS.

So i guess there''s a problem in result memory allocation, i
don''t know
what i could add to help you, i''m trying to reproduce it in a little 
code...

Ah, just another detail, this is how our index is created :
INDEX_OPTIONS = {
        :path => path,
        :auto_flush => false,
        :max_buffer_memory => 0x4000000,
        :max_buffered_docs => 100000,
        :use_compound_file => false}
INDEX = Ferret::Index::Index.new(INDEX_OPTIONS)

but i tried without any max_* and it''s the same...

Tell me what i can do to help if i don''t manage to reproduce it in a 
pastable code.

Thanks by advance,
Regards,
Jeremie ''ahFeel'' BORDIER

-- 
Posted via http://www.ruby-forum.com/.