Bug#701819: lintian: [reporting] Hard to interpret/ambiguous numbers

[Niels Thykier <niels@thykier.net>]

Package: lintian
Version: 2.5.11
Severity: minor

The front page on lintian.d.o has the following entries in its
"statistics" table.

"""
   Maintainers: 	2301 (+1)
   Source packages: 	17530 (-1)
   Binary packages: 	32288 (+3)
   μdeb packages: 	119 (+1)
"""

If you do not know the code, it is hard to tell that this is in fact
*NOT* the actual number of packages on lintian.d.o.
  No, these numbers are based on the packages that emit 1 (or more)
lintian tags.  For comparison, the actual number of source and binary
packages are closer to 20000 and 40000 (respectively) than the numbers
listed above.


Likewise, the (per-tag) graphs plots 3 numbers, "Tags", "Packages" and
"Overridden".  The actual meaning of these are left as an exercise to
the reader.
  One interpretation is that "Tags" is the number of times the tag is
seen (overridden or not) and "Overridden" is a the "subset" of these
tags that are overridden.  However, this is wrong[1].  The actual
meaning of these numbers are:

  Tags: Number of tags seen in the log that were not overridden
  Packages: The number of packages, which emitted this tag once (or more).
            Notable, this does not include any package in which all
            instances of this tag was overridden.
  Overridden: The number of overridden instances of this tag.

There is nothing wrong in these metrics themselves, but in their
current form people may misinterpret them and that is "bad"(tm).

~Niels

[1] Example where that interpretation does not hold includes:
  http://lintian.debian.org/tags/binary-or-shlib-defines-rpath.html