Subject: Re: Question on kernel building
To: Jason Thorpe <thorpej@nas.nasa.gov>
From: Chris G. Demetriou <cgd@cs.cmu.edu>
List: current-users
Date: 01/19/1997 15:42:54
> On Sat, 18 Jan 1997 10:40:13 -0800 (PST) 
>  "Gerald C. Simmons" <simmons@darykon.wa.com> wrote:
> 
>  > If I compile a kernel with "standard" serial port options, then add option
>  > COM_HAYESP in the config file, conf <machine>, make depend (in target
>  > directory,) then make, Why doesn't anything force the com.c driver to
>  > recompile? Seems I have to delete com.o inorder to force it to rebuild the
>  > driver. I thought "make depend" was suppost to cover this.
> 
> ...In order to do this, we'd have to emit a header file for every
> "well-known" option...  It would require some changes to config(8).
> 
> I've thought about doing this... and, since I've been hacking on
> config(8) for other reasons anyhow, I just might get around to
> this RSN.

While that's a good thing to do, i'd say it's completely wrong in this
instance.

using a #define to enable hayes ESP support is ... a hack, and a
broken one at that.  hayes ESP boards should be included via seperate
devices (sharing the back-end code that implements most of the driver,
of course).  The way the ESP probing code works in com.c is ... "very
wrong."


cgd