On 7/4/07 5:29 AM, Eric Botcazou wrote:
> /* Make sure result is not a complex expression consiting
> of operators of operators (IE (a + b) + (a + c))
> Otherwise, we will end up with unbounded expressions if
> fold does anything at all. */
> if (result)
> {
> if (is_gimple_min_invariant (result))
> return result;
>
> The expression
>
> &p__name_buffer[(<unnamed-signed:32>) MAX_EXPR <D.738_31, 0> + 1]
>
> is_gimple_min_invariant because it is TREE_INVARIANT. TREE_INVARIANT is
> correct because the semantics of Ada for IN parameters guarantees that the
> index is TREE_INVARIANT and p__name_buffer is a global variable.
The problem is that in GIMPLE we only allow TREE_INVARIANT as a gimple
value for ADDR_EXPRs. We still require a TREE_CONSTANT as an array
index. So, ARRAY[TREE_INVARIANT] is not valid GIMPLE.
FRE should refuse to substitute a TREE_INVARIANT expression as an
ARRAY_REF index.
> I think that the head comment is a little confusing: is it a formal predicate
> of the GIMPLE grammar, as most of its uses would suggest, e.g.
It is a formal predicate of the GIMPLE grammar. But there is an
inconsistency now. I don't think we want to allow min-invariant as
ARRAY_REF indexes because this would allow things like
&a[&b[3]]
which is too gross for words. So, the grammar should be updated to
include invariant addresses in 'val', but it should have a new rule so
that no invariant addresses are allowed as array indices.
