I brought this up a couple of weeks in the Lucene-user forum ("Out of
memory in Lucene 1.4.1 when re-indexing large number of documents") and
it started a thread that brought up some other out-of-memory problems.
Unfortunately the fixes for those introduced in 1.4.2 seem not to apply
to my problem:
When (re-)indexing a considerable number if objects running Lucene
version > 1.4RC3 in an IBM JDK 1.3.1 JVM the number of SegmentTermEnums
tends to mount up despite all forced gc runs until I get an
out-of-memory error. This is not so with a JDK 1.4.2 where it runs fine.
A quick fix that Doug Cutting proposed seems not to be working (version
1.4.1):
Index: src/java/org/apache/lucene/index/TermInfosReader.java
>>>>>>
===================================================================
>>>>>> RCS file:
>>>>>>
/home/cvs/jakarta-lucene/src/java/org/apache/lucene/index/TermInfosReade
r.java,v
>>>>>>
>>>>>> retrieving revision 1.9
>>>>>> diff -u -r1.9 TermInfosReader.java
>>>>>> --- src/java/org/apache/lucene/index/TermInfosReader.java 6
Aug
>>>>>> 2004 20:50:29 -0000 1.9
>>>>>> +++ src/java/org/apache/lucene/index/TermInfosReader.java 10
>>>>>> Sep 2004 17:46:47 -0000
>>>>>> @@ -45,6 +45,11 @@
>>>>>> readIndex();
>>>>>> }
>>>>>>
>>>>>> + protected final void finalize() {
>>>>>> + // patch for pre-1.4.2 JVMs, whose ThreadLocals leak
>>>>>> + enumerators.set(null);
>>>>>> + }
>>>>>> +
>>>>>> public int getSkipInterval() {
>>>>>> return origEnum.skipInterval;
>>>>>> }
>>>>>>
>>>>>>
>>>>>>
>>>>>>
------------------------------------------------------------------------
I did check this with Lucene 1.4.2 but the permanently increasing number
of SegmentTermEnum instances is still there with IBM JDK 1.3.1.
Daniel
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
