[d-kernel] [kaos@ocs.com.au: Re: integration support for 2.6.0 kernel in modutils]

[Gleb Stiblo <g.stiblo@sam-solutions.net>]

[-- Attachment #1: Type: text/plain, Size: 56 bytes --]

-- 
Gleb Stiblo AKA UlfR 	<g.stiblo@sam-solutions.net>


[-- Attachment #2: Type: message/rfc822, Size: 3147 bytes --]

From: Keith Owens <kaos@ocs.com.au>
To: Gleb Stiblo <g.stiblo@sam-solutions.net>
Subject: Re: integration support for 2.6.0 kernel in modutils
Date: Mon, 23 Feb 2004 19:58:58 +1100
Message-ID: <1694.1077526738@ocs3.ocs.com.au>

On Tue, 10 Feb 2004 13:11:09 +0200, 
Gleb Stiblo <g.stiblo@sam-solutions.net> wrote:
>KO> I have received your mail but I am on vacation until January 30.
>Hope your vacation was fine. So can we return to my question? :)

I had a look at your code over the weekend.  Sorry, but I do not want
to include kernel 2.6 support in modutils while keeping the same config
language.

The modutils 2.4 config language was never designed nor was it ever
well defined.  The "language" just grew as people came up with ideas.
There are nasty corner cases, recursion problems and features that
people rely on but those features only work by accident.  I could never
make changes to modutils config because it would break existing
behaviour that people were replying on.

I was planning to completely redo the modutils config language for 2.6
when Rusty came up with the in kernel linker.  To some extent I am glad
that 2.6 has simpler features, there is less to go wrong.

If you want to use some of the more complex features of modutils 2.4
configuration then the best approach is to extend mit with a well
designed and documented config language.  Make sure that people agree
what a featre should do before adding it.  Keeping the old and broken
2.4 config language will only propagate problems that I have been
trying to get rid of for years.