Re: [PATCH v7 4/8] sched: nominate preferred wakeup cpu

[Ingo Molnar]

* Vaidyanathan Srinivasan <svaidy@xxxxxxxxxxxxxxxxxx> wrote:

> * Vaidyanathan Srinivasan <svaidy@xxxxxxxxxxxxxxxxxx> [2008-12-20 16:06:02]:
> 
> > * Vaidyanathan Srinivasan <svaidy@xxxxxxxxxxxxxxxxxx> [2008-12-20 15:32:28]:
> > 
> > > * Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> [2008-12-19 20:44:55]:
> > > 
> > > > On Sat, 20 Dec 2008 10:06:38 +0530 Vaidyanathan Srinivasan <svaidy@xxxxxxxxxxxxxxxxxx> wrote:
> > > > 
> > > > > > 
> > > > > > kernel/sched.c: In function 'schedule':
> > > > > > kernel/sched.c:3679: warning: 'active_balance' may be used uninitialized in this function
> > > > > > 
> > > > > > This warning is correct - the code is buggy.
> > > > > 
> > > > > Yes this is my code bug.  I did not see the warning in sched.c.  Is
> > > > > there any build option that I need to pass in order to get -Wall
> > > > > effect?
> > > > 
> > > > That was just with plain old kbuild: `make allmodconfig;make'.
> > > > 
> > > > That warning was produced by gcc-4.0.2.  If you're using something more
> > > > recent then gcc has regressed.
> > > 
> > > This is an interesting problem.  I am unable to get that warning in the
> > > following GCC versions even with a combination of the following
> > > cmdline options: -Wall -Wextra -Wuninitialized
> > > 
> > > gcc version 4.1.2 20070502
> > > gcc version 4.2.3
> > > 
> > > I will look for older gcc and check this out.
> > 
> > I am able to get the above warning in 
> > gcc version 3.4.6 (Debian 3.4.6-9)
> > with default kbuild, no additional params.
> > 
> > Did not get the warning in 
> > gcc version 4.1.3 20080623
> > 
> > Hence there has been some change in GCC that prevented the
> > uninitialized warning.
> 
> Related GCC bug: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=179 Claimed 
> to have been fixed in GCC 4.4, but I don't know if we are ready to test 
> kernel compile with GCC 4.4.

4.4 is still way too noisy on x86 so not really practical.

> I have tested till gcc version 4.3.2 (Debian 4.3.2-1) and the bug is not 
> fixed.  I do not get the warning from sched.c.

neither did i get that warning. But Andrew did so thankfully, so we've got 
the fix :) We'll always need variance in testing.

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/