Re: [RFC: 2.6 patch] kernel/sys.c: possible cleanups

[Adrian Bunk <bunk@stusta.de>]

On Mon, May 01, 2006 at 02:07:22AM -0700, Andrew Morton wrote:
> Adrian Bunk <bunk@stusta.de> wrote:
> >
> > So the work would expand to:
> >  - writing 200 feature-removal-schedule.txt entries
> >  - marking 200 functions and variables as __deprecated_for_modules
> >
> >  And in a few months:
> >  - removing 200 feature-removal-schedule.txt entries
> >  - removing 200 __deprecated_for_modules markers
> >  - removing the 200 unused exports
> 
> Don't bother with all that stuff - a modprobe-time warning across a few
> kernel releases is sufficient to make any developers who are dependent upon
> an export aware of their problem.
> 
> Changing the export to EXPORT_UNUSED_SYMBOL() and then later removing the
> export is adequate.

OK, let's get this into 2.6.17 - it can't break anything and makes 
developers sooner aware of this.

Can we also create the rule that changing an EXPORT_UNUSED_SYMBOL() back 
to an EXPORT_SYMBOL() requires an in-kernel user of the export? 
Otherwise all this "there is no stable API for external modules" saying 
starts to become nonsense.

If we give developers 6 months of EXPORT_UNUSED_SYMBOL() warning this 
should be enough for them to submit their code for review and inclusion 
into the kernel.

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed