[parisc-linux-cvs] 2.6.0-test5-pa0 cleanups

[Grant Grundler]

Willy,
I'll commit the "Looks good" bits then...

On Thu, Sep 11, 2003 at 01:08:44PM +0100, Matthew Wilcox wrote:
> > drivers/serial/Kconfig
> > 	CONFIG_PDC_CONSOLE depends on CONFIG_VT in order
> > 	to pick up console_driver global.
> 
> Yes, but I think your patch to the comment is wrong.
...
> Enabling early console output is not done by CONFIG_PDC_CONSOLE but by
> changing
> #undef EARLY_BOOTUP_DEBUG
> to
> #define EARLY_BOOTUP_DEBUG

yes - another brain fart...not sure why I decided PDC_CONSOLE would
help me debug STI problems on c3k.

I'll fix the spelling error pointed out by Michael Wood and leave
the rest alone.

> > net/ipv4/ip_fragment.c
> > 	fix warning. atomic_sub() takes int for the first parameter.
> > 	"- sizeof()" is a long negative constant. Perhaps it's a
> > 	gcc 3.0.4 (hppa64) compiler bug that it doesn't get reduced to int.
> > 	No warnings for atomic_add(sizeof(foo),xxx) suggest it's normally ok.
> 
> Just remembering Rusty's comments about making interfaces intuitive to
> use,

Yeah - then should I file a gcc bug complaining the "- sizeof()"
expression should get reduced from "signed long" to "signed int" type?
(ie ip_fragment.c code "looks" right as is)

> how about the followig patch instead?

I'm afraid it will mask a case when someone really does use a
constant bigger than an int.
I'm not going to commit either my or your suggested patch.

...
> > @@ -291,7 +291,7 @@ static __inline__ int fls(int x)
> > -	"	zdep,TR		%0,29,30,%0\n"		/* y0000000 (y&3 = 0 */
> > +	"	zdep,TR		%0,29,30,%0\n"		/* y0000000 (y&3) = 0 */
> 
> I actually meant parentheses here, rather than operator precedence, so (y&3 = 0)

Ah ok. I'll commit that too.

thanks,
grant