Still catching up my emails.
On Mon, Jun 16, 2014 at 10:10 AM, Sam Ravnborg <sam@xxxxxxxxxxxx> wrote:
> sparse warned about "shift too big" if the rhs was a variable.
> As sparse usually cannot deduct the value of a variable
> silence these warnings.
That is very strange. It should not happen.
>
> Before this patch following code snippet would generate a warning:
>
> static int doshift(int foo)
> {
> return 1 << foo;
> }
>
> sparse did not know the value of 'foo' so reported a warning.
> Ignore the warning expect in the cases where the rhs of the shift
> is a constant.
That is strange. I can't duplicate the bug with your test case:
With tip of chrisl repository,
I get the following error with your test case:
./sparse /tmp/b.c
/tmp/b.c:3:18: warning: shift too big (1000) for type int
/tmp/b.c:8:22: warning: shift too big (1000) for type unsigned int
>
> diff --git a/expand.c b/expand.c
> index 0f6720c..ce955d3 100644
> --- a/expand.c
> +++ b/expand.c
> @@ -187,7 +187,7 @@ static int simplify_int_binop(struct expression *expr, struct symbol *ctype)
> return 0;
> r = right->value;
> if (expr->op == SPECIAL_LEFTSHIFT || expr->op == SPECIAL_RIGHTSHIFT) {
> - if (r >= ctype->bit_size) {
> + if (right->flags & Int_const_expr && r >= ctype->bit_size) {
This does not explain why it fix the bug. In simplify_int_binop()
There is the code check for this is constant expr on rhs. The non
constant value should just bail out early.
if (right->type != EXPR_VALUE)
return 0;
That is right before your change. Plus I can't duplicate the bug myself.
I suspect there is something else going on. Any way to help me reproduce
the bug?
I am using gcc 4.8.2 on x86_64.
Chris
--
To unsubscribe from this list: send the line "unsubscribe linux-sparse" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
