[PATCH 00/12] scsi/NCR5380: fix debugging macros and #include structure

[Joe Perches]

On Wed, 2014-03-19 at 12:46 +1100, Finn Thain wrote:
> On Tue, 18 Mar 2014, Joe Perches wrote:
> 
> > But using "if (0)" prevents the no_printk from occurring at all so there 
> > would be no side-effects and the format & args would still be verified 
> > by the compiler.
> 
> I'd prefer this (for symmetry and clarity):
> 
> #if NDEBUG
> #define dprintk(flg, fmt, ...) \
>         do { if ((NDEBUG) & (flg)) pr_debug(fmt, ## __VA_ARGS__); } while (0)
> #else
> #define dprintk(flg, fmt, ...) \
>         do { if (0) pr_debug(fmt, ## __VA_ARGS__); } while (0)
> #endif
> 
> But you seem to be asking for this instead:
> 
> #if NDEBUG
> #define dprintk(flg, fmt, ...) \
>         do { if ((NDEBUG) & (flg)) pr_debug(fmt, ## __VA_ARGS__); } while (0)
> #else
> #define dprintk(flg, fmt, ...) \
>         do { if (0) no_printk(fmt, ## __VA_ARGS__); } while (0)
> #endif
> 
> Why is that better?

It's not to me.

I suggested exactly your first block with if (0) pr_debug...
in the first thing I wrote.

https://lkml.org/lkml/2014/3/18/216

Geert suggested no_printk.

cheers, Joe