Unpredictable, but fairly frequent segfault using rb_eval

> I'm running a C++ app which calls rb_eval many times and perhaps once
> every 100 calls aborts due to segfault on rb_eval call
> 
> System is SUSE 9.3, behavior appears on 1.8.2 and 1.8.3 preview 1
> 
> I get the following:
> Program received signal SIGSEGV, Segmentation fault
> [Switching to Thread 1077648768 (LWP 8833)]
> 0x40306dac in _int_malloc () from /lib/tls/libc.so.6
> (gdb) bt
> #0  0x40306dac in _int_malloc () from /lib/tls/libc.so.6
> #1  0x40308344 in malloc () from /lib/tls/libc.so.6
> #2  0x08074a76 in ruby_xmalloc (size=4) at gc.c:121
> #3  0x080b642d in str_new (klass=Variable "klass" is not available.
> ) at string.c:92
> #4  0x080b6e2e in rb_str_new2 (ptr=0x82393f2 "2+3") at string.c:115
> #5  0x0806c490 in rb_eval_string (str=0x82393f2 "2+3") at eval.c:1555
> #6  0x0805e3ba in rb_protect (proc=0x806c460 <rb_eval_string>,
> data=136549362, state=0xbfffdf88) at eval.c:5256
> #7  0x0805e3e8 in rb_eval_string_protect (str=0x82393f2 "2+3",
> state=0xbfffdf88) at eval.c:1566
> 
> Any help would be greatly appreciated. System has 512Mb of memory, and my
> only thought is that this is an out-of-memory error. Btw, the eval'd
> expression is 2+3, not something complex.
> 
> Thanks,
> 
> Bill Barnhill