Subject: Re: Request for comments: let config(1) generate LKMs
To: Quentin Garnier <cube@cubidou.net>
From: Hiroyuki Bessho <bsh@NetBSD.org>
List: tech-toolchain
Date: 09/14/2007 02:15:56
At Thu, 13 Sep 2007 15:11:42 +0200,
Quentin Garnier wrote:
> 
> [1  <text/plain; us-ascii (quoted-printable)>]
> On Thu, Sep 13, 2007 at 06:55:32PM +0900, Hiroyuki Bessho wrote:
> [...]
> > I'd like some comments before I start further work on this.  
> 
> It looks strangely familiar to me:
> 
> http://mail-index.netbsd.org/tech-kern/2004/02/07/0001.html
> 
> Heh, config(1) was still config(8) at the time.
> 

  Oh, I've checked mail archives for prior art, but not down to 2004.
Actually, I've spent a-few-hours * a-few-days for this and came to
some point to get Makefiles for all modules.  So your patch is
strangely familiar to me :-)


> FWIW, I've killed that code off my trees a while ago.  I still think
> it's a good idea, but it certainly doesn't deal with all the issues
> involved when working with LKMs;  it's only a way to compile said
> modules.  My patch dealt with some level of dependency though.  I
> believe this is where config(1) has its saying, despite what Martin
> thinks:  using all the attributes defined in files.* is a very fine-
> grained but already existing tool to deal with module dependencies.
> 

  Your patch and the discussion on your proposal are very interesting.
I should study them if I go further.  Thank you for the info.

--
bsh.