Bug#750863: Please include package contents information

[josh@joshtriplett.org]

On Wed, Mar 18, 2015 at 06:36:02PM +0100, Lucas Nussbaum wrote:
> On 07/06/14 at 11:10 -0700, Josh Triplett wrote:
> > Package: qa.debian.org
> > Severity: wishlist
> > User: qa.debian.org@packages.debian.org
> > 
> > It'd be quite handy to have package contents information in UDD.
> > 
> > Columns, in addition to a unique package ID foreign-keyed to the
> > packages table:
> > 
> > filename
> > owner
> > group
> > mode
> 
> Hi Josh,
> 
> What is your use case for that?
> 
> This would add quite a lot of information to UDD, so I'd like to make
> sure it's really useful before implementing this.

Several use cases.  My original motivation in filing that bug is that
I'd like to search for files shipped in packages with user:group other
than root:root, or with modes other than 644 or 755 (as appropriate for
file/directory).  Right now, that would require downloading and
unpacking every package.

That actually suggests a useful compression technique: for owner, group,
and mode, you could provide a table or tables that *only* list unusual
files/directories that have user:group other than root:root or a mode
other than 644/755.  The rest of the data could then be synthesized with
an appropriate view.  That would eliminate the metadata entries for the
vast majority of the files in Debian packages, while transparently
providing all the same data.

I'd also like to search for packages that contain empty directories
(easily found by looking for directories without containing files) and
correlate that with whether any other package ships files in that
directory (or if it's provided solely for users).

And I'd like to search for things like filename conflicts and compare
them to package metadata; for instance, do packages that ship the same
filename conflict or not, and do the versions of the conflict match the
versions that ship the same file?  (That won't be perfect, since it
can't take diversions into account, which are unfortunately still
programmatic in maintainer scripts rather than being declarative.)

- Josh Triplett