On Sat, May 9, 2020 at 12:29 AM Bart Van Assche <bvanassche@xxxxxxx> wrote:
>
> >> Thanks for the feedback. I will have a look at _Pragma() and see what
> >> the best way is to suppress this warning.
> >
> > The __diag_ignore() macro in linux/compiler.h should work for this.
>
> Thanks Arnd, that's good to know. Is using __diag_ignore() mandatory in
> this case? The following construct seems to work fine with both gcc and
> clang:
>
> #define QLA_MSG_MAX 256
>
> +#pragma GCC diagnostic push
> +#ifndef __clang__
> +#pragma GCC diagnostic ignored "-Wsuggest-attribute=format"
> +#endif
> +
> DECLARE_EVENT_CLASS(qla_log_event,
> TP_PROTO(const char *buf,
> struct va_format *vaf),
> @@ -27,6 +32,8 @@ DECLARE_EVENT_CLASS(qla_log_event,
> TP_printk("%s %s", __get_str(buf), __get_str(msg))
> );
>
> +#pragma GCC diagnostic pop
> +
> DEFINE_EVENT(qla_log_event, ql_dbg_log,
> TP_PROTO(const char *buf, struct va_format *vaf),
> TP_ARGS(buf, vaf)
__diag_push(), __diag_ignore(), and __diag_pop() are just
shortcuts for open-coded #pragma plus #ifdef, they do
exactly the same thing here. I think it would be best to be
consistent and use the macros, but it works either way.
I actually have patches to introduce yet another syntax as part
of a larger rework, but that is still WiP.
Arnd
