On Tue, 31 Jul 2018 17:57:07 +0200 Paul Gevers wrote: > On 31-07-18 16:55, Chris Lamb wrote: > >> Rather unconstructively, they refuse to fix it until lintian has the > >> warning [2]. > > > > That seems.. strange. At the very least, Lintian is not Policy.
But lintian is the reason that the autopkgtests where added, since it complains when there are none. This is the simplest test that can be added, and I considered it better than none. It was initially suggested to help catch issues in the PGDG setup for the postgis package. > Ack. I filed 6 bugs today against packages in their team (the ones > currently trying to migrate). They already fixed the first bug. Fixed by removing the autopkgtests, since adding any is not worth the effort. > > Unrelated to the above, do you feel there any real fear that a /bin/ > > true would simply be replaced with something else if a maintainer saw > > this tag? There are, of course, plenty of no-op commands available in > > the base system... :) > > Yes, let's make sure the message is not "don't use /bin/true (or only > true, or exit 0 or echo)" but rather, please make sure the set of > autopkgtests actually test the installed package somehow. Note that testing the installed package "somehow" does not mean that the package is suitable for use. For example, just testing that the version the programs reports matches what's in debian/changelog doesn't mean that program actually works, it just didn't segfault for the --version commandline argument. Kind Regards, Bas
