Bug#683596: lintian: False positive "unused-override" when checking multiple packages with lintian
Control: tags -1 wontfix
On 2013-01-10 11:04, Niels Thykier wrote:
> On 2012-08-02 10:13, Matthijs Kooijman wrote:
>> Package: lintian
>> Version: 2.5.10
>> Severity: normal
>>
>> Hi folks,
>>
>
> Hi,
>
Hi again,
>> I just stumbled upon lintian giving me a false "unused-override"
>> tag. This happens when you override some tag because the file
>> needed for it is in another package, but then run lintian on both
>> packages involved: Lintian now finds the file in the other package,
>> doesn't trigger the warning and thinks the override is unused.
>>
>
> I have to admit I have been a bit conflicted in case. On one hand, I
> would like to promote the use of group processing precisely because it
> can give more accurate results.
> On the other hand, Lintian is now being used in "pre-install" checks
> (e.g. apt-daemon) where I am almost the checker will not fetch the
> related packages. Obviously these pre-install checkers will (depending
> on the tags emitted) refuse the install the package.
>
>> [...]
>> I'm not sure what the proper fix for this is. Perhaps each check should
>> detect that it's _not_ firing because there are multiple packages (and
>> it would have fired if the packages were processed separately), or
>> perhaps for each apparently unused override, the given check should be run
>> on each package separately to see if it's really unused?
>>
>
> If it is to be any of those two, I'd go for the former. Probably, the
> former is the only solution to this problem if we want to make packages
> Lintian clean in both use-cases.
>
> [...]
I have been thinking about this some more and I have come to the
conclusion that we cannot ensure packages are Lintian clean in "both
use-cases". For many checks, we might be able to do it, but there are
some where the output associated with the check depends on the packages
involved in the check.
Take "intra-source-package-circular-dependency" as an example. And that
(in theory) gets even more "fun" if you start processing binaries from
multiple architectures together (like we do on lintian.d.o). So at this
time, I am going to say "cantfix" to this.
~Niels
Reply to: