Bug 79875

We are getting errors with trace that contains unresolved symbols, e.g.:

==21510== 
==21510== 
==21510== 148 bytes in 1 blocks are definitely lost in loss record 32 of 131
==21510==    at 0x4002AC54: __builtin_new (vg_replace_malloc.c:172)
==21510==    by 0x4A724A38: ???
==21510==    by 0x4A6F873B: ???
==21510==    by 0x4A6F3BED: ???
==21510==    by 0x4A71851E: ???
==21510==    by 0x4A71756F: ???
==21510==    by 0x4058E790: Grape::CPT_TaskImp::execute(void) (/home/rotem-
g/rel080/grape/cpt/libcpt/CPT_TaskImp.cpp:1298)
==21510==    by 0x403D5761: Grape::GES_TaskThread::execute(Grape::GES_Job *) 
(/home/rotem-g/rel080/grape/ges/libges/SMP_Thread.cpp:804)
==21510==    by 0x403D62C0: Grape::GES_TaskThread::execContext(Grape::GES_Job 
*, Grape::CPT_Package &) (/home/rotem-
g/rel080/grape/ges/libges/SMP_Thread.cpp:903)
==21510==    by 0x403D6C98: Grape::GES_TaskThread::threadFunc(void) 
(/home/rotem-g/rel080/grape/ges/libges/SMP_Thread.cpp:986)
==21510==    by 0x406C71A2: DefThreadFunc(void *) (/home/rotem-
g/rel080/grape/osi/libosi/OSI_ThreadObj.cpp:18)
==21510==    by 0x406C6018: Grape::OSI_Thread::defThreadFunc(void *) 
(/home/rotem-g/rel080/grape/osi/libosi/OSI_PosixThread.cpp:86)
==21510==    by 0x406C5C49: DefaultThreadFunction (/home/rotem-
g/rel080/grape/osi/libosi/OSI_PosixThread.cpp:22)
==21510==    by 0x4029F923: thread_wrapper (vg_libpthread.c:667)
==21510==    by 0x401735F3: 
(within /product/pci/sw/local/lib/valgrind/valgrind.so)
==21510== 
==21510== 

This phenomena makes us unable to trace back to the leak position.
In the same run we get many good reports tracing the stack without any 
unresolved symbols.
Types of reports for which we saw this phenomena were "definitely lost" 
and "still reachable" error reports.

Our executable uses dynamic libraries (loaded explicitely by the application).

system information:
gcc version 2.96 20000731 (Red Hat Linux 7.2 2.96-112.7.2)

Valgrind used:
We used valgrind 2.0.0. 
Also tried to run valgrind 2.1.0 and 2.1.1 but they crushed immediately.