Fw: [cgl_discussion] Sources from each build...

[Peter Badovinatz]

Craig Griffin wrote:
> 
> Khalid Aziz wrote:
> >
> > By publishing the source code from last known good build, we give people
> > a convenient way of downloading the last known good sources as a single
> > file. In general, when one checks out the top of tree from any CVS
> > archive (CGL or not), one does not expect it to be in usable shape at
> > all times. Top of the tree may have bugs, or might not build even.
> > Sources from the last known good build are the only way to get source
> > code you can put some faith in. Of course, what would be more reliable
> > would be the sources from last completely tested and validated set of
> > bits, but we are not there yet. Downloading a single tar file is much
> > more convenient than checking sources out of CVS. The audience for this
> > source tarball would be open source community and distros, primarily.
> 
> And, a single tar file is exactly what MontaVista needs to be able
> to integrate and productize from CGL sources in an effective way.

I don't say this lightly, but, is this single tar file approach
appropriate for all other parties that may desire to make use of this
code?  I see the argument for posting only the tarball as being
reasonable because that simply shows the differences and we shouldn't do
anything further than that.

> 
> In fact, we need a single tar file of the sources you used to build
> and we need it for EVERY build you post, so we can diff from build
> to build and so we can follow the threads of integration.

The true source of code for open source projects should be those
project's home project sites, should it not?  For instance, any
distribution not based on kernel 2.4.18 but still wishing to fulfill
OSDL CGL requiremets is not going to be able to make direct use of these
bits and will need to go back to the original sources anyway.  Most
projects will continue their development apart from CGLE and satisfy
CGLE as but one of any number of interested integrators.

And given the dynamism of the Linux kernel, and timing of distribution
releases, kernel skew will be a reality.  CGLE is a point-in-time
integration.

What they get from CGLE is an understanding that the pieces have been
put together and an understanding of the level of work to do so.

> 
> Trying to wade through all the CVS stuff, from an MV integration
> standpoint, from build to build, would be a nightmare.

MV is one valued and interested integrator.  IMO CGLE has to be a
resource that draws many parties to appreciate OSDL CGL if we want
complete acceptance.  It has to show that our Architecture, Requirements
and Compliance documents are grounded in something resembling reality.

I know we don't have a compliance document, but we've been told we have
to have one.

> 
> Output whatever format you think necessary to validation/QA.
> --
> Craig Griffin, QA Engineer
> MontaVista Software Inc.
> 1237 E. Arques Avenue
> Sunnyvale, CA 94085
> (408) 992-4720 Direct Number
> (408) 835-9972 Cell
> cgriffin at mvista.com

Peter
--
Peter R. Badovinatz aka 'Wombat' -- IBM Linux Technology Center
preferred: tabmowzo at us.ibm.com / alternate: wombat at us.ibm.com
These are my opinions and absolutely not official opinions of IBM, Corp.