"Hosts" within tinderbox?

[Ade Lovett]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


On Oct 06, 2007, at 02:31 , Erwin Lansing wrote:
> I don't think this is widely used, if at all.  I tried to use it  
> for the
> shared tinderboxes, but failed miserably (I can't remember the  
> details,
> might have been something with different src trees on different
> architectures), so I ended up having two indepedent webinterfaces,  
> only
> the portstrees/ directory is shared.  It would be a great feature  
> if it
> was fixed though :-)

Therein lies my hidden agenda, as it were.

Having done plenty of investigation on this subject, I have come to  
the conclusion that in order to support multiple build hosts etc., is  
going to require essentially a ground-up rewrite.  We've tried to  
shoehorn in concepts, but the overall infrastructure just simply  
isn't designed for multiple systems.

So, what I'm proposing is as follows:

1.  A cleanout of detritus within tinderbox-HEAD (such as removing  
Hosts).  Overall, the code in HEAD is considerably cleaner than 2.x,  
and this will also help in terms of not having to manage two  
tinderbox branches.

2.  A new release (based on HEAD) of the tinderbox code, with the  
TINDERBOX_2_X branch being EOL'd.  Whether we call it 3.0, or 2.6, or  
Depleted Uranium, I really don't care.  The point being that we move  
to a single tinderbox code, which snapshot releases on the way for  
inclusion in the ports tree.

3.  A sit down whiteboard session, identifying requirements for a  
*new* tinderbox system (multi host, multi arch, and so on), with a  
ground up rewrite.

Comments as always.

- -aDe

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iD8DBQFHB22rpXS8U0IvffwRAsOWAJ47alvGQrG+cp9Spfl9N9sJjqLOqQCfZE5G
n9WUU9igtsC9+pDuCWncoA0=
=f7Bk
-----END PGP SIGNATURE-----