On Fri, Jun 10, 2016 at 02:36:23PM +0200, Marek Polacek wrote:
> --- gcc/gimplify.c
> +++ gcc/gimplify.c
> @@ -1559,6 +1559,50 @@ gimplify_statement_list (tree *expr_p, gimple_seq
> *pre_p)
> return GS_ALL_DONE;
> }
>
> +/* Possibly warn about unreachable statements between switch's controlling
> + expression and the first case. SEQ is the body of a switch expression.
> */
> +
> +static void
> +maybe_warn_switch_unreachable (gimple_seq seq)
> +{
> + if (!warn_switch_unreachable
> + /* This warning doesn't play well with Fortran when optimizations
> + are on. */
> + || lang_GNU_Fortran ()
> + || seq == NULL)
> + return;
> +
> + /* Look into the innermost lexical scope. */
> + while (seq && gimple_code (seq) == GIMPLE_BIND)
> + seq = gimple_bind_body (as_a <gbind *> (seq));
> +
> + /* If it is empty, there's nothing to do. */
> + if (seq == NULL)
> + return;
Won't this just give up on say:
void
foo (int a, int b)
{
switch (a)
{
{ int c; }
{ int d; }
{ int e; }
b++;
case 1:
break;
}
}
? Such blocks can also be can be arbitrarily nested.
Wouldn't it be better to just walk_gimple_seq with NULL callback_op
and non-NULL callback_stmt that would stop on the first real stmt in there?
Also, the above loop looks confusing, I'd be expecting gimple_seq_first_stmt
(seq) before testing it for GIMPLE_BIND or casting to gbind.
Jakub
