[pylucene-dev] pylucene and 2gb limit of files

[Andi Vajda]

On Thu, 3 Aug 2006, tsuraan at tsuraan.net wrote:

>> Quite possibly. Does 'make test' succeed ?
>
> Not entirely; everything works except for test_PythonDirectory and 
> test_RangeFilter, which both segmentation fault.  The stack trace goes like 
> this:
>
> #0  0xb7d48adf in sigprocmask () from /lib/tls/libc.so.6
> #1  0xb6d3ce38 in _Jv_SetInitialHeapSize ()
>   from /usr/lib/gcc/i686-pc-linux-gnu/4.1.1/libgcj.so.7
> #2  0xb6d3d5e0 in _Jv_ThrowBadArrayIndex ()
>   from /usr/lib/gcc/i686-pc-linux-gnu/4.1.1/libgcj.so.7
> #3  <signal handler called>
> #4  0xb7368a00 in GC_local_gcj_malloc ()
>   from /usr/lib/gcc/i686-pc-linux-gnu/4.1.1/libgcj.so.7
> #5  0xb6d3d3ca in _Jv_AllocObjectNoFinalizer ()
>   from /usr/lib/gcc/i686-pc-linux-gnu/4.1.1/libgcj.so.7
> #6  0xb6d3d436 in _Jv_AllocObject ()
>   from /usr/lib/gcc/i686-pc-linux-gnu/4.1.1/libgcj.so.7
> #7  0xb6d3d5f9 in _Jv_ThrowBadArrayIndex ()
>   from /usr/lib/gcc/i686-pc-linux-gnu/4.1.1/libgcj.so.7

Looks like you're on Linux ? I thought you were on freebsd (pardon the 
confusion, I'm on vacation at the moment and not following too closely).

With the patch mentioned here:
   http://gcc.gnu.org/ml/java/2006-03/msg00190.html
I've been able to use gcj 4.1.0 on gentoo. Note that this was 5 months ago and 
I can't verify this until I'm back in the office in a few weeks.
Before using this patch, see if bug 13212 solves this as well since it seems 
to supercede the patch.

> And then lines 3-7 repeat forever, same thing for both tests.  I've never 
> seen such a deep stack trace.  Anyhow, I guess I'll complain to the gcj 
> people if the linked patch doesn't fist this, and if I can find a better way 
> to report the error.  Do you have any idea why PyLucene's make test isn't 
> part of the standard testing procedure for the gcj team?  It seems like one 
> of the better tests of whether the compiler is functional in a given 
> configuration.

That is a great idea. The gcj team hasn't shown much interest yet in using 
PyLucene to reproduce problems with gcj. Another idea that was suggested 
recently on java-dev at lucene.apache.org is that Java Lucene officially support 
gcj. They asked me to try to get their test suite (much more complete than 
PyLucene's which is a partial port of Java Lucene unit tests and samples) to 
run with gcj. Java Lucene has more visibility than PyLucene and doesn't 
involve python and that would make it easier, maybe, for the gcj developers to 
use it for reproducing and investigating gcj-related Lucene gcj compilation 
problems.

I expect to get into this next month...

In the meantime, if you can get the gcj developers' interest tickled with the 
PyLucene unit tests, that would be great. Whenever the problem has been simple 
enough to reproduce, they've been very helpful !

Andi..