Re: [PATCH] init: Fix printk format strings

[Randy Dunlap <randy.dunlap@oracle.com>]

On Thu, 11 Oct 2007 18:09:47 +0200 Vegard Nossum wrote:

> On 10/11/07, Randy Dunlap <randy.dunlap@oracle.com> wrote:
> > On Thu, 11 Oct 2007 17:55:16 +0200 Vegard Nossum wrote:
> >
> > > On 10/11/07, Randy Dunlap <randy.dunlap@oracle.com> wrote:
> > > > On Thu, 11 Oct 2007 08:17:02 +0200 Vegard Nossum wrote:
> > > >
> > > > > This makes sure printk format strings are string literals containing no
> > > > > more than a single line.
> > > >
> > > > Each patch needs justification (unless it is blatantly obvious).
> > > >
> > > >
> > > > > Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
> > > > > ---
> > > > >  init/calibrate.c        |    4 +++-
> > > > >  init/do_mounts_initrd.c |    5 ++++-
> > > > >  init/main.c             |    2 +-
> > > > >  3 files changed, 8 insertions(+), 3 deletions(-)
> > > > >
> > > > > diff --git a/init/do_mounts_initrd.c b/init/do_mounts_initrd.c
> > > > > index fd4fc12..ad6174c 100644
> > > > > --- a/init/do_mounts_initrd.c
> > > > > +++ b/init/do_mounts_initrd.c
> > > > > @@ -98,7 +98,10 @@ static void __init handle_initrd(void)
> > > > >                       error = sys_ioctl(fd, BLKFLSBUF, 0);
> > > > >                       sys_close(fd);
> > > > >               }
> > > > > -             printk(!error ? "okay\n" : "failed\n");
> > > > > +             if(error)
> > > > > +                     printk("failed\n");
> > > > > +             else
> > > > > +                     printk("okay\n");
> > > > >       }
> > > > >  }
> > > >
> > > > Why this one?
> > > > and if it must change, use:
> > > >                 if (error)
> > > > but I see little need for the change.
> > >
> > > Oops about the space. And it's to make the format string a constant
> > > literal. If it isn't, you are going to run into trouble with options
> > > to remove printks at compile-time based on log-level token. I realize
> > > it's not really an issue at this point in time, but I assume that it
> > > will eventually make its way into the kernel in some way or another.
> >
> > so would this be OK?
> >
> >                 printk("%s\n", error ? "failed" : "okay");
> 
> It helps the filtering, but it doesn't help localisation very much, I
> think. Since for that, you'd only translate the format string, which
> in this case isn't much to translate. (I do realize that this is a
> very, very tiny part of the kernel and probably hundreds of other
> call-sites have much worse problems -- but every little helps, and now
> that you're at it... :-))

Oh well.

I find the proposed changes too restrictive, practically changing
a stream-oriented output into a variable-length block-oriented output.


---
~Randy