[PATCH] Handle read-only vars initialized to string literals in c_strlen

[Richard Kenner]

    > Why?  It seems to me that the lower-level we do such a check, the better.

    It is not a check, it is constant folding.  The higher-level we
    expose constants, the better.

It's checking for a constant.  I agree it should be exposed as early
as possible, but the routine that checks for it should be as low as possible
so that as many callers of it will see it.