[coreboot] Warnings

[Myles Watson]

On Fri, Apr 30, 2010 at 11:50 AM, Stefan Reinauer <stepan at coresystems.de> wrote:
> On 4/30/10 7:13 PM, Myles Watson wrote:
>
> As of 5512, these are the last four warnings that are not "defined but
> not used" warnings.
>
>
> What's your impression... should we activate -Wno-unused ? or should we
> attempt to fix them?

I don't think it's worth activating -Wno-unused-functions untill we
activate -Werror, but I think we should do that soon so that we don't
reintroduce a lot of warnings.

> I'd like to go drop all / most of the #ifdef UNUSED_CODE segments from
> coreboot over time, making it easier to follow the code that is actually
> there.
I agree.

> Most of the existing warnings are amd k8 / fam10 warnings... and it's kind
> of tricky to understand why one board does init different than another one.
> Why don't all boards use " wait_all_aps_started" and / or "
> wait_all_core0_started" and / or "allow_all_aps_stop"? I think the unused
> messages show that we have too much unnecessary diversity here (aka rank
> growth)
Yes.  There are definitely some things that need to be standardized,
and the warnings are good at pointing that out, as you've said.

> src/northbridge/amd/gx2/chipsetinit.c:271: warning: suggest
> parentheses around '-' inside '<<'
>
>
> This would need help from someone with a GX2 (or willing to check out the
> data sheets ;-)
>
> src/northbridge/amd/lx/raminit.c:334: warning: array subscript is
> above array bounds
>
>
> Marc indicated, this could be a compiler "problem". We could maybe add an
> explicit check for the array index.

It looks like it's being indexed by a function that's built into the
compiler that returns bit positions.  Since the return value could be
up to 31, I think it's a correct warning (the array's only got 8
elements.)

> src/northbridge/via/vx800/pci_rawops.h:43:2: warning: #warning "FIXME:
> get rid of this extra copy of pci access functions."
>
> Gone with 5513
>
> src/southbridge/intel/pxhd/pxhd_bridge.c:70:2: warning: #warning
> "Please review lots of dead code here."
>
> Gone with 5514

It's much nicer to have so many fewer warnings!

Thanks,
Myles