Re: Introducting Debian Trends: historical graphs about Debian packaging practices, and "packages smells"

[Niels Thykier <niels@thykier.net>]

Andreas Tille:
> Hi Niels,
> 
> On Tue, Apr 16, 2019 at 12:54:00PM +0000, Niels Thykier wrote:
>>> speaking about false positives:
>>>
>>>        libhmsbeagle (U)     should switch to dh. Current build system: debhelper (source version: 3.1.2+dfsg-5)
>>>
>>> I have no idea what might have triggered this on the current d/rules
>>> file[1].
>>>
>>> Kind regards
>>>
>>>        Andreas.
>>>
>>> [1] https://salsa.debian.org/med-team/libhmsbeagle/blob/master/debian/rules
>>
>> Try:
>>
>>   lintian  -L =classification <your-packages-here>
>>
>> And you will probably see something like:
>>
>>   C: <pkg> source: debian-build-system other
>>
>> I.e. most likely a false-positive in lintian.  Once confirmed, please
>> file a bug against lintian or/and provide a MR on salsa for fixing the bug.
>>
>> For reference, given that trends.d.n relies solely on a lintian (with a
>> handful of patches to add a few tags), the issues are most likely in
>> lintian or in the particular patches (for the latter, please follow up
>> on the relevant bugs filed by Lucas against lintian to add those tags).
> 
> As far as I understood Lucas is talking about a patched lintian and
> 
>   grep -R "should switch to dh" /usr/share/lintian/*
> 
> has no results so this warning is not part of lintian. 

Hi Andreas,

While Lucas is talking about a patched lintian, the concrete case is
almost certainly a part of "standard" lintian.  I remember implementing
the tag to detect the "debian build system" and I pretty sure it does
*not* cope with "VAR=VALUE dh" calls.

You are right that "should switch to dh" does not occur in the lintian
codebase.  This is because lintian does not assign this value to the
missing of a classification tag - this is where Lucas code takes the
lintian output and translates the result to "package smells".  As a part
of this, he basically translates (among other):

  C: ...: debian-build-system X

into

  ... should switch to dh. Current build system: X (...)

When X != 'dh'.

> Similarly
> 
>     prottest (U)         should switch to dh. Current build system: debhelper (source version: 3.4.2+dfsg-3)
>     python-pyfaidx (U)   should switch to dh. Current build system: debhelper (source version: 0.5.5.2-1)
> 
> are false positives of this lintian fork.
> 
> Kind regards
> 
>       Andreas.
> 

They might be a false-positive, but I am pretty sure this particular bug
is not in Lucas's fork.  To my knowledge, that patch simply adds 4
classification tags that have been posted to the lintian BTS and those
have no conditional logic of their own (they emit tags via existing
branches inside lintian AFAIR).

Thanks,
~Niels