Re: voyager_{thread,cat}.c compile warnings

["Cédric Augonnet" <cedric.augonnet@gmail.com>]

2007/7/22, James Bottomley <James.Bottomley@hansenpartnership.com>:
> On Sun, 2007-07-22 at 18:49 -0400, Cédric Augonnet wrote:
> > iff -urN a/arch/i386/mach-voyager/voyager_cat.c
> > b/arch/i386/mach-voyager/voyager_cat.c
> > --- /home/gonnet/tmp/linux-2.6.22/arch/i386/mach-voyager/voyager_cat.c  2007-07-20 11:50:17.000000000 -0400
> > +++ linux-2.6.22/arch/i386/mach-voyager/voyager_cat.c   2007-07-22
> > 11:24:34.000000000 -0400
> > @@ -682,7 +682,7 @@
> >                         outb(VOYAGER_CAT_END, CAT_CMD);
> >                         continue;
> >                 }
> > -               if(eprom_size > sizeof(eprom_buf)) {
> > +               if((unsigned)eprom_size > sizeof(eprom_buf)) {
>
> Actually, no.  If gcc can deduce that the comparison is always false
> then I want it not to build the body of the if.  The only thing I don't
> know how to do is to shut up the warning in this case.  What you've done
> is make gcc pretend it doesn't know the if is always false.
>
> >                         printk("**WARNING**: Voyager insufficient size
> > to read EPROM data, module 0x%x.  Need %d\n", i, eprom_size);
> >                         outb(VOYAGER_CAT_END, CAT_CMD);
> >                         continue;
> > @@ -752,7 +752,7 @@
> >                         outb(VOYAGER_CAT_END, CAT_CMD);
> >                         continue;
> >                 }
> > -               if(eprom_size > sizeof(eprom_buf)) {
> > +               if((unsigned)eprom_size > sizeof(eprom_buf)) {
> >                         printk("**WARNING**: Voyager insufficient size
> > to read EPROM data, module 0x%x.  Need %d\n", i, eprom_size);
> >                         outb(VOYAGER_CAT_END, CAT_CMD);
> >                         continue;
> > diff -urN a/arch/i386/mach-voyager/voyager_thread.c
> > b/arch/i386/mach-voyager/voyager_thread.c
> > --- /home/gonnet/tmp/linux-2.6.22/arch/i386/mach-voyager/voyager_thread.c       2007-07-20 11:50:17.000000000 -0400
> > +++
> > linux-2.6.22/arch/i386/mach-voyager/voyager_thread.c        2007-07-22
> > 11:27:13.000000000 -0400
> > @@ -92,7 +92,7 @@
> >         }
> >  }
> >
> > -static int
> > +static void
> >  thread(void *unused)
> >  {
> >         printk(KERN_NOTICE "Voyager starting monitor thread\n");
>
> You didn't actually compile this, did you?  Apparently the signature of
> the kthread_run function changed from returning void to returning int.
> Unfortunately the person who fixed this up forgot to add a return 0 at
> the end of the voyager thread() function .. which is the correct fix.

Arg i was caught by that one.

> James
>

Ouch indeed this quick'n'dirty patch was, let's call it a full mistake
:) sorry for that, it could indeed not be tested as i don't have the
hardware.

Still, is it safe to compare two variable with different types anyway ?

In http://lists.infradead.org/pipermail/linux-pcmcia/2004-March/000586.html
they also have the same issue, they just do
s/ foo > 0xffff / foo & ~0xffff /
should not it solve the problem as well ?

Sorry again for the first patch, next time i'll just shut up.

Regards,
Cédric