[pylucene-dev] Segfault in JCCEnv::deleteGlobalRef

[Christian Heimes]

Hello!

JCC and PyLucene are causing a segfault on a busy machine several times
a day. Using a debug build of Python and PyLucene I was able to narrow
the cause of the bug down. So far I've examined six core dumps. The seg
fault is always caused by the deallocation of an IndexSearcher instance.

#7  <signal handler called>
#8  0xb162fb05 in JNIEnv_::DeleteGlobalRef (this=0x0, gref=0x95bf2e0)
    at /opt/vlsdbg/lib/jdk1.6/include/jni.h:830
#9  0xb162f1bf in JCCEnv::deleteGlobalRef (this=0xb20a6d78,
obj=0x95bf2e0, id=30740434)
    at build/_lucene/JCCEnv.cpp:204
#10 0xb139fa20 in JObject::operator= (this=0x51c8cc08, obj=@0x68dfd0a4)
    at JCC-2.1-py2.5-linux-i686.egg/jcc/sources/JObject.h:75
#11 0xb160dbe9 in t_JObject_dealloc (self=0x51c8cbf8) at
build/_lucene/JObject.cpp:100
#12 0xb7ea7aaf in _Py_Dealloc (op=0x51c8cbf8) at Objects/object.c:1933

(gdb) print op->ob_type->tp_name
$4 = 0xb1647152 "IndexSearcher"

Python: 2.5.2
JCC: 2.1
PyLucene: 2.4.0.2
Java: Sun Java 1.6.0 update10
gcc version 4.1.2 20070115 (SUSE Linux)

Christian