The lli program fails on the following instruction when the -force-interpreter
option is used:

br bool setne (void ()* %_GLOBAL__D_lshaped_driver.cpp_00000000_A3849043, void
()* null), label %then.0.i, label %loopentry.i

The lli program aborts with:

Unhandled ConstantExpr:  bool setne (void ()*
%_GLOBAL__D_lshaped_driver.cpp_00000000_A3849043, void ()* null)


Examination and some extra debugging added to
lib/ExecutionEngine/Interpreter/Execution.cpp seems to indicate that the setne
instruction is castable as a constant expression (which makes sense because null
is constant and the %_GLOBAL... is the name of an internal function).  However,
the interpreter cannot interpret a setne constant expression, so lli aborts.

I'll try to create a smaller test case later tonight.

This patch adds support for setne constant expressions (among others):

http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20031208/010072.html

Thanks,
-Brian