Re: [PATCH][tuples][Ada] Disable gcc_asserts in gimple_has_side_effects

[Diego Novillo <dnovillo at google dot com>]

On 7/20/08 11:34 AM, Richard Guenther wrote:
> With Ada we hit these asserts a lot, and I don't see why they are correct
> in any case.

Because if any operand has side-effects, the optimizers shouldn't be 
doing anything to them, so we want them marked as having volatile operands.

Do you have any counter example in mind? (a side-effects operand that 
should not mark the statement as having volatile operands).


> Still, the one case I looked at looks like superfluous
> TREE_SIDE_EFFECTS from the Ada side:
>
>   foo = {}

This does seem superfluous.  An empty constructor shouldn't have 
side-effects.


Diego.