On Wed, Nov 15, 2017 at 10:10:59AM -0800, Russ Allbery wrote:
>
> Distribution packages generated by upstream are usually horrible unless
> upstream is deeply involved in that distribution community. From the
> perspective of an experienced packager for that distribution, they are
> usually way behind best practices, don't use common facilities, install
> into weird locations, and otherwise look like something that someone just
> beat on with a hammer until it vaguely installed and sort of worked.
I think Steffen's point was that all the hideousness you are talking
about was solved in version a.b.c of the software and if version
a.b.(c+1) builds and passes our test suite it will most probably not
have changed.
So for packages without lots of rdepends I think the approach to
auto-update is worth some testing ... provided somebody will do the work
to setup such a system which is most probably the cruxial point. For
instance currently CRAN packages are auto-build - I would not see any
big issues to auto-update r-cran-* packages.
Kind regards
Andreas.
--
http://fam-tille.de
