Re: [PATCH 1/2] Relocator framework

[Vladimir 'phcoder' Serbinenko]

On Sat, Aug 8, 2009 at 5:52 AM, Pavel Roskin<address@hidden> wrote:
> On Fri, 2009-08-07 at 14:01 +0200, Vladimir 'phcoder' Serbinenko wrote:
>
>> Apple's compiler is based GCC but binutils aren't and they pose the
>> most of problems. Actualy the most problematic bit was that I didn't
>> know that unless you prefix variable with L_ apple's assembler treats
>> it as global. Adding L_ in this code solves problem. Yves proposed to
>> add a LOCAL(x) macro to symbol.h this way the assembler code is also
>> more readable. I also did further adjustments in this patch to
>> decrease the number of ifdefs but had no time to test it and so
>> couldn't submit it
>
> I would prefer if you don't commit any patches that introduce any more
> preprocessor conditionals for the Apple compiler or assembler.
LOCAL(x) can always be expanded to L_## x but it will make code more
readable and modifiable in future. And it will allow a bunch of other
ifdef's go out.
>
> It would make it hard to debug problems reported by others, as different
> compilers would produce different binaries.
>
Even just changing compiler options you produce different binary
> If we are using some feature that is just nice but not needed (such as
> binary & in lnxboot.S to set the image size to 1024), it would be better
> to avoid using that feature for all compilers rather than for the
> compilers that don't support it.
>
lnxboot.S with Apple's toolchain is disabled temporarily awaiting
someone to find a definitive solution.
Why do our 2 new maintainers propose to shave off useful features?
> --
> Regards,
> Pavel Roskin
>
>
> _______________________________________________
> Grub-devel mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/grub-devel
>



-- 
Regards
Vladimir 'phcoder' Serbinenko

Personal git repository: http://repo.or.cz/w/grub2/phcoder.git