Hi Felix, thanks for your reply e for taking care of package shape via Lintian. I much appreciated your huge restyling activity and your tons of commits addressed to make Lintian's internals more performant and consistent.
First of all I have to make clear that my observations come from my *user* point of view, I'm pretty sure that a plenty of useful statistics can be pulled out for a Debian Developer [1] or other users. Il 04/06/20 20:24, Felix Lechner ha scritto: > Your request predates my involvement but I would like to understand > how to make our data more meaningful on an archive level. (The > 'archive' is actually a fluid set of packages that is constantly > synchronized via dakweb.) I can find meaning only by slicing the data > in other dimensions, such as by architecture, or by providing > normalized averages, such as tags per package or overrides per > package, and so on. Definitively per-tag statistics. > Below you will find an old snapshot from the Internet Archive's > Wayback Machine. Which data was most useful to you, please, and why? > > Kind regards, > Felix Lechner > > * * * > > Archives > > The following archives are processed by Lintian: > > Archive nameAttributeAttribute value > debianArchitecturesi386 amd64 > Distributions/Suitesunstable experimental > Componentsmain contrib non-free > Mirror timestampSun, 21 Apr 2019 20:30:22 +0000 > debian-debugArchitecturesi386 amd64 > Distributions/Suitesunstable-debug experimental-debug > Componentsmain contrib non-free > Mirror timestampSun, 21 Apr 2019 20:30:22 +0000 Doesn't matter archive witch stats come from, nor archs involved. Aggregate them is a perfect solution for me. As side note, experimental repository, from my point of view, have little interest for users. Given ephemeral nature of stuff in experimental, I think users are not motivated to report bug against this kind of packages nor make aware developers about problems in files of functionalities that may disappear the day after. Anyway, my usual workflow was to follow "List of emitted tags grouped by severity" link, than I used to check "error" tags or tags emitted for spotting errors in package descriptions. Sometimes tag names didn't have an immediate meaning but their web pages are explanatory enough. So nevermind. > > Statistics > > Last updated: Mon, 22 Apr 2019 00:33:21 +0000 > Maintainers: 2423 (+0) > Package groups: 31284 (+5) > Rescheduled groups: 3 (+0) > Groups with processing errors: 3 (+0) I think these stats are for the sake of Lintian Developers... I never knew what those meant :) > > Source packages: 29929 (+1) > Binary packages: 39928 (-3) > μdeb packages: 225 (+0) > E Errors: 32305 (+4) > W Warnings: 186976 (-6) > I Info tags: 575740 (-196) > P Pedantic tags: 175454 (-42) All of these are very useful. > > O Overridden tags: 181682 (-138) Overrides represents stat makes me put in anger the most, and I consider them an easy way to bypass problems instead of fixing them. I've always thought overrides should tend to 0 in a reasonable time, but I see their number is ever growing. Maybe Lintian have to drop/revisit tags that are more overridden? Or avoid overrides for a specific tag subset? Like the previous one, this aspect definitively does not belong to users as well. > X Experimental tags: 124653 (+25) Useful Sorry for digressing in some point from your request, I hope Lintian web pages become a useful instrument for Debian users and Developers. Deb packages needs to be functional, compliant and.... nice to see :) Only this matters. Thanks again for your work. P.S. Can you restore on the mainpage the link pointing to the git repo? [1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=945544
