[pylucene-dev] optimize in 2.0.0

[Andi Vajda]

On Thu, 13 Jul 2006, Yura Smolsky wrote:

>>> gcj (GCC) 4.0.3
>
> AV> What OS ? If Linux, which one ?
>
> Debian Linux 2.6.11 #1 SMP Tue Apr 12 10:08:12 CDT 2005 i686 GNU/Linux
> we have built 1.4.3 successfully before.
>
> 'make test' returns this:
>
> [... skipped ...]
> ======================================================================
> ERROR: testGetBestFragmentsSimpleQuery (__main__.HighlighterTestCase)
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>  File "test/test_Highlighter.py", line 81, in testGetBestFragmentsSimpleQuery
>    self.doStandardHighlights()
>  File "test/test_Highlighter.py", line 103, in doStandardHighlights
>    highlighter.setTextFragmenter(SimpleFragmenter(20))
> JavaError: java.lang.NullPointerException
... skipped ...

Well, it looks like either your gcj is not producing sane PyLucene executables
or that the libgcj used is not sane. Verify that your LD_LIBRARY_PATH is 
indeed picking up the libgcj you built with your gcj 4.0.3 build.
A Suse Linux user just solved a similar problem by fixing his LD_LIBRARY_PATH.

If that doesn't solve it, try older gcj versions. 3.4.6 should work quite well 
except for the 2 Gb file size limit bug fixed in gcj 4. I realize this may not 
work for you since you have huge indexes. In this case, try newer versions of 
gcj, such as gcj 4.1.1 or even 4.2.0.

If that still doesn't solve it, track the actual problem down (by stepping 
through Java Lucene's compiled code in gdb) and report it to java at gcc.gnu.org, 
if you provide them a reproducible test case, they're quite diligent in fixing 
it or in providing you with a way around it.

Andi..