Fix processing of ADDR_EXPR in get_expr_operands

[Diego Novillo]

On Thu, 2004-07-29 at 22:38, Richard Kenner wrote:

>     Invariant addresses are supposed to be marked with TREE_INVARIANT, not
>     TREE_CONSTANT.  That's what I've been told by the FE folks and that's
>     why the tree optimizers almost always rely on
>     is_gimple_min_invariant().
> 
> What does this have to do with the volatile question I'm asking about?
> 
It's rather obvious.  You cannot rely on TREE_CONSTANT to determine
constness of ADDR_EXPRs.

I don't know *why* ADDR_EXPR of a volatile is not TREE_CONSTANT.  I
don't really care.  All I know, and always use, is that constness of
ADDR_EXPRs is tested with TREE_INVARIANT.  Hence the call to
is_gimple_min_invariant.  Even TREE_CONSTANT is not always right, you
have to check for overflow as well.

> these flags on ADDR_EXPRs and they would be blown away by the middle-end
> if it did.  They are set by recompute_tree_invarant_for_addr_expr.
>
Which you will notice sets both TREE_CONSTANT and TREE_INVARIANT, not
necessarily with the same value.


Diego.