This message might be of interest to other @rpm5.org based developers/distros as well.

Mancoosi is a EU research project looking at the scalability
of package management in FLOSS distros.

CUDF is a markup that is being used as input to a SAT (SATisfiability see wikipedia)
approach to depsolving dependency assertions (aka Mancoosi Work Project 5 WP5).

What is unique with Mancoosi WP5 CUDF is that both apt/rpm package problems
can be represented.

And what will be very useful imho (once there are some well understood
problems represented in CUDF) for @rpm5.org development is the ability
to easily generate test cases to ensure no unexpected behaviors.

73 de Jeff

Begin forwarded message:

> From: Jeff Johnson <n3npq@mac.com>
> Date: January 8, 2010 11:57:31 AM EST
> To: mancoosi-wp5@sympa.mancoosi.univ-paris-diderot.fr
> Bcc: Anders F Björklund <afb@algonet.se>
> Subject: Generating RPM test packages from CUDF
> 
> Anders Bjorklund (Smart maintainer, also @rpm5.org) just
> shared a perl cudf2cpec script with me.
> 
> The cudf2spec perl script transofrms CUDF into a *.spec
> \recipe that can be fed to rpmbuild to generate binary packages.
> 
> The discussions about RPM Conflicts: and Macriva DUDF -> CUDF
> have indicated a new usage case for CUDF, generating test
> packages for existing RPM implementations, in order to
> sort out CUDF <-> RPM discrepancies.
> 
> Something as simple as the attached script gets most of the job done.
> 
> (aside)
> I have a "cudftool" executable using lubcudf-0.5.94 that I
> will likely enhance with the ability to generate slightly
> differently formatted spec files, where 3 sets of subpkgs
> for
> 	installed
> 	available
> 	problem
> can be built from 3, rather than zillions, of *.spec files.
> 
> cudftool will be in rpm-5.3 when released.
> 
> The immediate usage case (@rpm5.org) will be generating
> "make check" fodder for unit test coverage.
> 
> There's a longer term usage case for CUDF in RPM as well, but CUDF (and WP5)
> has chosen to type
> 	Version: 1
> as an integer, and the needed memoization to convert back to
> integers preserving order will need some further work.
> 
> Mancoosi/EDOS test cases are already quite useful @rpm5.org.
> 
> E.g. This LCOV/GCOV link here
> 	http://jbj.org/files/rpm/rpm-5.3/lcov
> you all know as the
> 	EDOS car-glass problem.
> Quite a bit of RPM code gets exercised with even simple
> problems like car-glass even with no files in the
> generated toy/test *.rpm packages.
> 
> hth
> 
> 73 de Jeff

>