Messages in this thread | | | Date | Mon, 22 Dec 2008 15:13:38 -0800 | From | Andrew Morton <> | Subject | Re: [patch] powerpc: change u64/s64 to a long long integer type |
| |
On Tue, 23 Dec 2008 00:00:35 +0100 Sam Ravnborg <sam@ravnborg.org> wrote:
> On Mon, Dec 22, 2008 at 02:43:19PM -0800, Andrew Morton wrote: > > On Mon, 22 Dec 2008 09:03:41 +0100 > > Ingo Molnar <mingo@elte.hu> wrote: > > > > > Subject: powerpc: change u64/s64 to a long long integer type > > > > <applause> > > > > There are lots of other architctures that need doing though. > > I have missed the introduction. > Can you explain why?
People keep on doing
printk("%llu", some_u64);
testing it only on x86_64 and this generates a warning storm on powerpc, sparc64, etc. Because they use `long', not `long long'.
And not just a little bit - this bug gets repeated maybe ten times per week - it's insane.
If we make all architectures use `long long' then the above code becomes correct and warning-free on all architectures.
And we then get to remove all the open-coded (unsigned long long) casts which we added all over the tree (hundreds of them).
> I have done a bit of sparc hacking lately and was wondering if > sparc needs something similar. > > cd arch/sparc > git grep u64 | grep -v __u64 | wc -l > 448 > > [On the unified tree]
Yes, quite a few 64-bit architectures are using `long' for their s64/u64 types. We should convert them all to `long long'.
That's quite trivial to do, but it causes a lot of warnings due to code in arch-speciic files which does
printk("%lu", some_u64);
which will then generate a warning. Or an error if the archtiecture uses -Werror, which several do. So we need to simultaneously convert all those to %llu. (Or %Lu, which saves a byte :))
| |