Files conflicts in ports

[Kubilay Kocak]

On 10/26/17 11:00 PM, Rodrigo Osorio wrote:
> 
> On 10/26/17 13:45, Kubilay Kocak wrote:
>> On 10/26/17 3:58 PM, L.Bartoletti wrote:
>>> Hi Rodrigo,
>>>
>>> Thank you for this precious tool.
>>>
>>> One question, seeing one of my ports which have conflicts
>>> (devel/py-gtfslib
>>> http://pkgtool.osorio.me/conflicts/lbartoletti@tuxfamily.org.html). Is
>>> it or not good to install test files?
>>
>> They're all effectively upstream bugs: installing modules into shared
>> locations. 'tests' is a common enough module name that its one of the
>> most easily observed in practice.
>>
>> There's nothing intrinsically wrong with tests being installed, but they
>> should be under/within their package module directories.
>>
>> Most projects exclude them (from installation) with something like:
>>
>> packages = find_packages(exclude=[...]),
>>
>> Though doing the above for a project with this packaging 'bug' is not
>> really the correct solution. Maybe for a short term
>> files/patch-setup.py, but report it upstream
>>
>>> Regards.
>>>
>>> Loïc
>>>
>>> On 10.10.2017 20:52, Rodrigo Osorio wrote:
>>>> Dear port maintainers,
>>>>
>>>> It appears that a number of ports install files with the same names at
>>>> the same locations,
>>>> causing file conflicts and unexpected behaviors for users.
>>>>
>>>> To help solving this issue I ran a tool to list per maintainer the
>>>> conflicting ports with
>>>> the list of impacted files ; the list is updated every day at 4am UTC.
>>>>
>>>> http://pkgtool.osorio.me/conflicts/
>>>>
>>>> I believe most of the conflicts are trivial and can be solved with a
>>>> proper declaration in the CONFLICTS variable.
>>>> So take a look at it and don't hesitate to come back to me if you have
>>>> questions.
>>>>
>>>> best regards,
>>>>
>>>> - rodrigo
>>>>
> I agree with Kubilay, If tests aren't relevant for production use the
> can be skipped.
> The point here is many (if not all) py- packages install the same test
> files and this is wrong.
> 
> - rodrigo

Just to be explicit, in describing them as upstream bugs, I didn't also
mean they're *not* port bugs.

The above ports, and any port in fact, that currently install
conflicting files, must either:

- Add CONFLICTS[_*] with all of their conflicting ports, OR
- Not install them

This is separate from the issue of
value-of-installed-python-tests-for-*package*-users (*ports* users can
run them via the sdist in WRKSRC), and separate from the method of
resolving the conflict (removal, rename, upstream bug fix, patch, etc)

./koobs