svn commit: r334593 - in head/dns: bind96 bind96/files bind98 bind98/files bind99 bind99/files

[John Marino]

On 11/26/2013 16:48, Eygene Ryabinkin wrote:
> Tue, Nov 26, 2013 at 08:46:47AM +0100, John Marino wrote:
>> On 11/26/2013 04:53, Alexey Dokuchaev wrote:
>>> On Sun, Nov 24, 2013 at 10:37:01AM +0100, John Marino wrote:
>>>> I have attached a diff with the proposed changes covering all three
>>>> ports.  It's pretty straight-forward.
>>>>  
>>>>  .if ${PORT_OPTIONS:MREPLACE_BASE}
>>>> -.if ${OSVERSION} >= 1000500
>>>> +.if ${OPSYS} == FreeBSD && ${OSVERSION} >= 1000500
>>>
>>> Given that OSVERSION is used extensively in the tree, I'm worrying that
>>> adding these extra "${OPSYS} == FreeBSD" checks will our makefiles less
>>> readable.  Wouldn't it make more sense to place this check somewhere in
>>> single place, and define whatever OSVERSION that looks reasonable when
>>> OPSYS == FreeBSD instead of patching every port?
>>
>> There is a "discussion" about this on the portmgr list.  Well, it's not
>> really a discussion.  I started it, bapt answered it once, and the
>> portmgr stopped talking about it.
>>
>> As far as "less readable", that really should not be a concern at all.
>> It is simply invalid to use OSVERSION without OPSYS since it's OPSYS
>> specific.
> 
> What are the other OPSYS that FreeBSD ports collection should support?
> 
> If there are any that are really needed, then OPSYS must be combined
> with OSVERSION, but if there is no strong current demand for that, it
> is probably not the right time to add complexity to the Makefiles.

*Hundreds* of problems with ports have already been detected and fixed
as a direct result of the creation in dports.  In many cases, problems
were fixed on F10 before the general users saw them.  Based on that
track record alone, does that not merit using OSVERSION properly to
support another system that brings those benefits?  And yes, it's
possible that yet other platforms can use ports in the future.  Right
now each new instance of OSVERSION can break a port, so now is the right
time to not only make developers aware of that but to reducing its use.

I am also not agreeing that prefixing OSVERSION with "OPSYS == <value>
&&" is "complex".  People seriously are not going to have trouble with
the concept that OSVERSION is platform specific.

John