Luca Falavigna <[email protected]> writes: > I recently conducted a quick analysis to spot empty packages in the > archive [1], some of these packages were fixed or will be in a short > time frame, so I'm happy it helped maintainers to fix grave bugs.
> I was suggested to implement a lintian check for this issue [2], so I'm > asking you how feasible my approach could be, or if you have better > solutions to achieve this goal. Here is how I got my results [3], and > how I'd like to implement such check. > A package is declared empty if all of this conditions are met: > * package does not ship files outside of /usr/share/doc/$pkg > * package does not have subdirectories in its /usr/share/doc dir > * package does not have files in its /usr/share/doc dir but common > ones (copyright, changelog*, README, AUTHORS, NEWS, ...) > * package does not have a "whitelist" word in its description: > - meta > - transition > - dummy > - dependency package > - empty package > - virtual package This looks like a good approach to me. There's a similar test in Lintian already to determine which packages are metapackages, but it's currently used in a much more limited way. I think Lintian excludes anything that's named like a documentation package, but I like your approach better. > I'm aware this could lead to false positives. In my test, preliminary > list of packages included some packages which didn't declare themselves > as "meta packages", but I think they should warn users about their > "meta" status, so they can be eventually removed. Probably would be good to accompany this with a patch to devref recommending a way to denote such packages. -- Russ Allbery ([email protected]) <http://www.eyrie.org/~eagle/> -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
