Crash in codegen of switch with overflow

switch (1) {
case 111111111111111111111111111111111111111:
;
}

generates invalid bitcode with clang -emit-llvm. Looks like codegen
isn't using the truncated value that sema is calculating. (Maybe we
should be putting an ImplicitCastExpr in here?)

-Eli

Yep, you're right. Fixed:
http://lists.cs.uiuc.edu/pipermail/cfe-commits/Week-of-Mon-20080114/003814.html

Thanks!

-Chris