Bug#415558: lintian: too pedantic about /usr/lib vs. /usr/share (image-file-in-usr-lib)
Package: lintian
Version: 1.23.27
Severity: wishlist
lintian warns in a couple of places about files being placed in /usr/lib
vs. /usr/share. While I appreciate that the FHS says "The /usr/share
hierarchy is for all read-only architecture independent data files", I
nevertheless think that these warnings are too pedantic.
Consider the intent of /usr/share. An administrator of many systems of
heterogeneous architectures is intended to be able to share the
/usr/share directory among them, for example using NFS. If one places
architecture-dependent files in /usr/share, then clearly that is a bug
because it certainly defeats that purpose. However, placing an
architecture-independent file in /usr/lib or elsewhere merely wastes a
little disk space on such systems. In some cases (e.g. menu files, small
images, etc.), the amount of disk space involved is truly negligible in
comparison to the amount of disk space required for the entire
installation. In many cases, it is not at all clear to me that it is
truly worth the developer time and potential for introducing new bugs
required to move the files.
Don't get me wrong: I'm not saying that we should fail to fix bugs just
because "it's too much effort". Nor am I saying that the effort expended
so far on moving files to /usr/share is wasted; hierarchies such as
/usr/share/doc, /usr/share/fonts, and /usr/share/man are a very
substantial win on heterogeneous shared systems. However, I do think
that we're rapidly approaching diminishing returns on this. When the
lintian warning is for a single image file in /usr/lib that requires
fiddly build system changes to move (or fragile hacks in debian/ that
would be liable to break on each new upstream version), I question
whether this is a sensible use of developer time, and I question whether
it really buys anything for administrators of heterogeneous systems.
Most of the warnings in question are simple transitions in progress that
are largely handled by a single debhelper command (menu-file-in-usr-lib)
or are confined to a small group of packages
(package-installs-nonbinary-perl-in-usr-lib-perl5). I'm not so bothered
about those, since the practices for dealing with them are pretty
established now.
However, I do think image-file-in-usr-lib ought to be downgraded to an
informative message. It's the sort of thing that bites in all sorts of
random packages (see
http://lintian.debian.org/reports/Timage-file-in-usr-lib.html) and is
typically fiddly and varied to deal with, and a lot of the images
involved are little things like icons where the effort doesn't really
justify the gain; if they were embedded into executables, we likely
wouldn't care.
It's been a while since I was particularly involved in lintian
maintenance, so I don't feel that I can change this unilaterally. What
do the other lintian maintainers think of my position?
Thanks,
--
Colin Watson [cjwatson@debian.org]
Reply to: