ancient-standards-version and "replay-ability"

[Niels Thykier <niels@thykier.net>]

Hi

We have a mantra[1] with Lintian that goes something like this "If you
run lintian on the same package twice, you should get the same result".
 The one exception stated so far has been "unless you upgraded or
downgraded lintian or its dependencies".

However, I feel "ancient-standards-version" violates this and I believe
we should fix this special case.  The violation arrives from the
following code snippet in checks/standards-version:

    if ($obsdate + (60 * 60 * 24 * 365 * 2) < time) {
        tag 'ancient-standards-version', $tag;
    } else {
       [...]

Where we compare the release date of a given standards-version
($obsdate) to the system time.


I see two problems issues with having this special case; the first is
that "we say one thing, but we do another".  Obviously relying on the
system-time is less likely to give weird results than (i.e.) using the
APT cache, but I still feel we are not following our own "rule" here.

The second issue is that breaks our test-suite every 2nd year (a.k.a. at
least once within the average stable release).  I have a feeling that a
number of "Debianites" would consider this an RC-bug.
  This issue can be solved partly using date + S-V substitutions in the
t/source framework (which I have been working towards), but it will not
cover the "out-of-date-standards-version" test(s).  Those /could/ be
excluded from the build-time tests, but I would prefer we grew a better
solution here.

~Niels

[1] I say mantra because I could find it in the User Manual, but it
shows up in various bug replies, over IRC etc.  I might have missed it
in the manual... whatever.