paul <goodoldpaul@autistici.org> writes:

> Hi Marius,
>
> On 6/22/20 10:16 PM, Marius Bakke wrote:
>> Giacomo Leidi <goodoldpaul@autistici.org> writes:
>>
>>> * gnu/packages/syncthing.scm (syncthing-gtk): New variable.
>> [...]
>>    
>>> +    (build-system python-build-system)
>>> +    (arguments
>>> +     `(#:python ,python-2))
>>> +    (propagated-inputs
>>> +     `(("python2-bcrypt" ,python2-bcrypt)
>>> +       ("python2-dateutil" ,python2-dateutil)
>>> +       ("python2-pycairo" ,python2-pycairo)
>>> +       ("python2-pygobject" ,python2-pygobject)
>>> +       ("python-nautilus" ,python-nautilus)
>>> +       ("libappindicator" ,libappindicator)
>>> +       ("libnotify" ,libnotify)
>>> +       ("psmisc" ,psmisc)
>>> +       ("syncthing" ,syncthing)))
>> I don't think these needs to be propagated, as the 'syncthing-gtk'
>> executable will be automatically wrapped with a PYTHONPATH that contains
>> the Python inputs.
> I didn't know that, I really should study deeper the build systems.
>> I'm not sure about libappindicator, libnotify and psmisc though.
>> Ideally we'd insert absolute references where appropriate to avoid
>> propagation.  In the worst case we can wrap it with those in PATH.
>>
>> Can you look into it and send an updated patch?
>
> As you suggested I managed to move all propagated-inputs to the inputs 
> field, I'm not sure I did it right but from my tests the package seem to 
> be working.

Excellent.

> I'm attaching an updated version of the patch, I apologize for the delay.

Apologies in return, this got lost in my email queue.

Applied!