[PATCH] variable size arrays in nested functions

[Richard Kenner]

    It's used in situations where we really want to re-evaluate everything
    in subexpressions, such as when an expression is used both in the main
    body of the function and in an EH cleanup.

That's what I thought.  That is indeed, a real kludge.

But why would this be easier if we replaced the SAVE_EXPR with an unnamed
VAR_DECL and its initalization?  Wouldn't it be *harder* to do the
reevaluation there because we'd have to reallocate the DECL_RTL?