Re: [PATCH] [0/6] kfifo fixes/improvements

[Stefani Seibold <stefani@seibold.net>]

Am Montag, den 28.12.2009, 02:41 +0100 schrieb Andi Kleen:
> On Sun, Dec 27, 2009 at 04:12:06PM -0800, Roland Dreier wrote:
> > 
> >  > I am not happy to see you to take over my project. Especial as most of
> >  > your fixes are part of my new macro based implementation. Have a look at
> >  > http://patchwork.kernel.org/patch/69093/
> > 
> > I don't really understand.  You spent a lot of time getting the kfifo
> > stuff merged, and now you want to merge (quoting from that patch above)
> > "a complete reimplementation of the new kfifo API"?
> > 

Yes, because of the limitations. The new merge kfifo stuff was based on
the old one. So i overtake this it. But the new one is fully compatible
to the merged kfifo.

> > What happened here?  Couldn't you have done the reimplementation before
> > merging?
> 

I am sorry, but did not recognized all constrains and features which are
really necessary for a real generic fifo interface. And also i did't saw
the possibility to do it as a template, because C does not support it.
It takes time the mature the idea to implement this as a macro set.

BTW, you give me the idea to reimplementation for kfifo, because you ask
me if it is not possible to merge my kqueue RFC. 

> I guess the reimplementation came too late (happens sometimes)
> And I agree that making kfifos record oriented makes sense.

What does it mean? To late for 2.6.33 or to late to replace it for ever?
I think it is easy to replace, because it is fully tested and 100
percent compatible to the new kfifo implementation.

> 
> Still now that the old one is in we have to fix it at least
> until there are no users left.
> 

The only user of the new features are currently you.