i was trying to get a new directory implementation going based on the
pythondirectory support in pylucene. i ran into some problems, and was
curious if this part of the codebase is in use, i couldnt find any unit
tests for it, nor references of folks using it.
after indexing about 20 files, i get an io error
Traceback (most recent call last):
File "manidx.py", line 103, in ?
indexDirectory(path)
File "manidx.py", line 43, in indexDirectory
indexFile(dir, name)
File "manidx.py", line 85, in indexFile
writer.addDocument(doc)
File "/usr/lib/python2.4/site-packages/PyLucene.py", line 1902, in
addDocument
def addDocument(*args): return _PyLucene.IndexWriter_addDocument(*args)
PyLucene.JavaError: java.io.IOException: read past EOF
not really sure, what the problem, its not calling back into python
afaics when the error happens.
probably unrelated but the operations i see on the impl just before the
exception
dir.lock commit.lock
lock.obtaintimeout 10000
dir.create segments.new
file.close segments.new
dir.rename segments.new segments
dir.exist deletable
dir.open deletable
file.close deletable
lock.release
which was the same sequence after the first 10, which went without
incident. there haven't been any reads api calls during the indexing.
are there some rules to reference counting with objects being used by
java? is there a way to trackdown/pinpoint this sort of error better?
platform ubuntu breezy - ppc - gcj 4.0.1 - pylucene svntrunk - py2.4.2
thanks
kapil
_______________________________________________
pylucene-dev mailing list
[email protected]
http://lists.osafoundation.org/mailman/listinfo/pylucene-dev
