[Bf-committers] New project, let's move on!

John K. Walton bf-committers@blender.org
Sun, 15 Dec 2002 20:54:09 -0500 (EST)


On Sun, 15 Dec 2002, Hans Lambermont wrote:

> John K. Walton wrote:
> 
> > I think by and large most people want to work on blender code, 
> > and are picking the cvs01 or independent cvs projects to work on.
> > so the blender code base is fragmenting anyway. if it has to come
> > together sometime, might as well do it now:
> > 
> > 1) move back to old cvs (or make new one) this one it the closest to
> >    being crossplatform (as much as NaN was anyway)
> > 2) overlay autoconf efforts to develope in parallel.
> > 3) afterward being consolidation of cvs01, and independant code fixes
> >    to get back to 'one for all, and all for one code tree'
> 
> Interesting approach, I have no problems with it, but I doubt if many
> others think the same ;-) In any case I'm convinced we need tinderbox
> auto-compilations, so that we continuously know the build status on all
> 6 platforms. And it's an excellent way to expand to other platforms
> (Solaris/i386, BeOS/i386, Linux/ppc, Linux/alpha, OpenBSD, which all
> share the honour of having Blender ported  to them once ;-)

i agree, there has to be a way to schedule regressions in a predicable
fashion (once we start rolling).

Is it possible for me to have one of my machines (or a number of
them) available at my house belong to the tinderbox compile farm?
I have cable modem.
 
> I'm quite busy nowadays with my son Fyor :) 


ha! kids are a lot of work, no?  :)

> so I hope someone else can
> help me set set up tinderbox on the blender.org website. I made an
> install document when I set it up at NaN, see
> http://lambermont.webhop.org/snippets/tinderbox.txt and the newer
> tinderbox client perl script is in
> intern/tools/tinderbox/tinderbox-blender.pl so things should be fairly
> easy.


I would love to help, but it would be like 'the blind leading the blind'
since I have high maintenance kids too :
 
> > otherwise, we need to make fixing (did it ever work) the cvs makefiles
> > in the cvs01 tree a priority.
> 
> The NaN Makefiles in cvs01 are almost the same as what we used at NaN.
> Stuff that was changed was about the removal of some code that was not
> to be open-sourced, that's all.

hrm, it failed for me last night, however the autoconf make didn't (well 
it did, i got a segfault).

> Difficulties arise mostly because the lib/ and extern/ directories were
> not released (not sure about licensing, that's why I didn't put them in).
> 
> I think I need not explain that the NaN makefiles worked on all 8
> platforms and that we used them for everything, including our releases
> (which I still *oops* need to add to the cvs01 repository). The
> tinderbox client system even built the complete .tar.gz releases (except
> for the auto-installer on windows) that were used on the NaN ftp and web
> server every 15 minutes (but we didn't upload them every 15 minutes ;-).
> 
> What about creating a shell script per platform that sets some default
> environment variables for that platform and calls (h)make, like Willem
> van Engen did at : http://www.stack.nl/~wvengen/compileblender That
> would get things compiled from scratch more easily.


anything, and everything helps atm.