[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: NM's to fix bugs? (Was: Asking for an advocate (gURLChecker))



On Sat, 12 Apr 2003, Jarno Elonen wrote:

> Since stable base packages (in addition to effective developmen tools!) are 
> the most important factor in making a stable OS, it would be good to have 
> more work concentrated on them.

As Colin has already pointed out, bugs.qa.debian.org can give lists of
base/essential/etc packages which have bugs.

> Maybe some semi-automatic (eh..) list of priorized packages would help? 
> Something like:
> 
>   package-use-frequency * (num_of_grave_bugs*1.0 + num_of_critical_bugs*0.5)
>   * number of dependencies/2 + subjective_bias_from_DD_vote_or_something

package-use-frequency can be impossible to work out, unless you work from
the popularity-contest data, which cannot be taken as definitive
(interesting though it is).

Rather than number of number-of-dependencies, I'd work on reverse
dependencies, that is, the number of packages which depend on this one. 
Also work on a dep tree, so that even though only one package might depend
on yours, 40 packages might depend on that one other package.  Have a rating
system, perhaps, which halves the importance for each level you move away
from the package in question.  Have a max of perhaps 3 or 4 levels. 
Circular dependency checking would also need to be added.

Sounds like something a prospective DD might like to have a go at. 
Especially someone who has plans to perhaps hack dpkg, apt, or something
like that.  Groping through the packages DB is the main part, counting
outstanding bugs should be pretty simple.  Anyone want to give it a whirl?

- Matt




Reply to: