Re: [PATCH] drivers/hwmon: Use pr_fmt and pr_<level>

[Jean Delvare <khali@linux-fr.org>]

On Wed, 20 Oct 2010 12:52:03 -0700, Joe Perches wrote:
> On Wed, 2010-10-20 at 21:38 +0200, Jean Delvare wrote:
> > On Wed, 20 Oct 2010 16:19:46 -0300, Davidlohr Bueso wrote:
> > > Well we do have this:
> > > #ifndef pr_fmt
> > > #define pr_fmt(fmt) fmt
> > > #endif
> > I can read the code, thank you. My point is that what we have is
> > insufficient, as Joe is proposing to use a different pr_fmt definition
> > for pretty much all drivers in the kernel tree.
> 
> No, I'm proposing to use a uniform pr_fmt definition for
> files that want a prefix on their logging messages.

Which files can legitimately not want a prefix on their logging
messages? How many files does this represent?

> At some point, the files that do not want a prefix
> should be converted to use #define pr_fmt(fmt) fmt.
> 
> Repeating from another email, there are ~1500 files that
> use pr_<level> without using pr_fmt.  There are 335 files
> that #define pr_fmt.
> 
> So I'm not saying that point in time is now.

I am saying it is now.

> Another way to do this would be to standardize
> hwmon.h so it would add:
> 	"#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt"
> in only 1 place.
> 
> So perhaps add this to include/linux/hwmon.h instead:
> 
> #if defined pr_fmt
> #undef pr_fmt
> #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
> #endif

And do the same in all other subsystems? What's the benefit over doing
the same in <linux/kernel.h>.

As it stands, I simply don't want your patches. As Guenter rightly
said, these are cleanup patches which don't make the code cleaner, so
there's no point. If you want to clean this up, please do it correctly.

-- 
Jean Delvare