Summary: experiences with NanoBSD, successes and nits on a Soekris 4801

[Poul-Henning Kamp]

In message <m2acl36jum.wl%gnn at neville-neil.com>, gnn at freebsd.org writes:
>At Mon, 04 Jul 2005 10:25:34 +0200,
>Poul-Henning Kamp wrote:
>> I played with developing that graph by removing lines from LINT
>> and see what compiled and what didn't.  Based on the progress I
>> made I would estimate the full graph will take about 1 CPU-year to
>> calculate by trial&error.
>
>Hmm.  Well, either I need a very fast CPU, or a more clever approach.
>I guess we'll see...

This was a dual Opteron, 2GHz, 4GB.

It might be possible to do a less brute force and more analytical
approach:

Start out with sys/conf/files*, then add sys/conf/options.

Then for each and every .c file, grep out the #includes and
build a tree of complete _potential_ dependencies.

Also grep out all #ifmumble constructs and record those.

You now have a huge data structure from which it should be possible
to determine potential dependencies.

Ie:  Which source files could possibly be affected by this option
or conversely, which options could possibly affect this source file.

By pruning the trival cases from the data structure, the brute force
work would be a lot less.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.