Re: [PATCH 01/12] making the kernel -Wshadow clean - fix mconf

[Andrew Morton <akpm@osdl.org>]

On Sun, 30 Jul 2006 21:17:18 +0200
"Jesper Juhl" <jesper.juhl@gmail.com> wrote:

> > (looks at
> > lock_cpu_hotplug())
> >
> Hmm, I'll take a look at lock_cpu_hotplug() - can you provide
> additional details?
> 

eh.  We put the recursive-sem thing in there as a temp fix to cpufreq to
get 2.6.something out the door, expressing fine intentions to fix it for
real later on.  Then look what happened.  Don't go there.

> 
> > That being said, no, we can't go and rename up().  Let us go through the
> > patches one-at-a-time..
> >
> i figured as much. *But* won't you agree that up_sem() (or considering
> the other locking function names, sem_up() would probably be better)
> would be a much better name for a global like this one?
> 
> How about a plan like this:
> We introduce sem_up() and sem_down() wrapper functions now. They could
> go into 2.6.19 for example - and also add a note to
> feature-removal-schedule.txt that the old function names will be
> removed in 1 year. Then in a few kernel versions - say 2.6.21 - we
> deprecate the old names and add a big fac comment in the source as
> well.
> Wouldn't that be doable?   Or do we have to live with up()/down() forever?

Well actually when struct mutex went in we decided to remove all
non-counting uses of semaphores kernel-wide, migrating them to mutexes. 
Then to remove all the arch-specific semaphore implementations and
implement an arch-neutral version.  After that has been done would be an
appropriate time to rename things.

But it never happened.  See "fine intentions", above ;)