Current problem reports assigned to you

[Alexander Nedotsukov]

Jeremy Messenger wrote:

>>> > o [2004/06/15] ports/67970 gnome       ports textproc/libxml,
>>> > textproc/libxslt:
>>>
>>> Well.......I personal don't care about this since pkgconfig is mofo  
>>> small.
>>> :-P I bet this one is going to take forever unless someone step in and
>>> create the patch(es).
>>
>>
>> I don't see why we need to get anymore clever with pkgconfig.  It is
>> small, and is generally needed on any desktop system.
>
>
> In case if anyone want to know my real thought of this PR. I disagree 
> with  this PR, because the purpose of install foo.pc is to run 
> 'pkg-config  --options foo'. Therefore, pkgconfig should be install.

    eik@ insist that he strictly follow the Porters Handbok (which is 
good thing by itself imho). And I bet he will respond you Jeremy with a 
sample like the only purpose to install .h files is to complille them. 
Why not to depend on gcc then? Or even better the only purpose to 
install .la files is to use them with libtool so why not to run-time 
depend on it? Btw this PR was colsed once. And I wish you good luck ;-)
    If it does matter my understanding of problem is:
- there is no clear definition of what library run-time dependency is [1]
- there is no clear direction on how to handle cases when port 
originally installs more that one entity (in our case library binary + 
pkgconfig metainfo) [2]
- and we can not cleanup pkgconfig directory w/o adding it to mtree or 
depend on some dedicated like gnomehier port.
    The later is the only real reason imho we run-time depend on 
pkgconfig ATM. All others just an hidden excuse of this fact.

[1] Here is the source of all those speculations who better unredtands 
Porters Handbook statements.
[2] This may become more serious than it is now if some day we will be 
required to run pkg-config --install libfoo.pc

All the best,
Alexander.