PYTHON_VERSION=python3.4 and USES=python:2,build

[Marcus von Appen]

"John W. O'Brien" <john at saltant.com>:

> On 12/1/14 4:01 AM, Marcus von Appen wrote:
>>
>> Zitat von Andrew Berg <aberg010 at my.hennepintech.edu>:
>>
>>> On 2014.11.30 08:23, John W. O'Brien wrote:
>>>> On 11/29/14 11:40 PM, Andrew Berg wrote:
>>>>> On 2014.11.29 22:00, John W. O'Brien wrote:
>>>>>> Let's say that I want to build net/py-pyzmq with
>>>>>> PYTHON_VERSION=python3.4.
>>>>> Why this instead of using DEFAULT_VERSIONS?
>>>>
>>>> I was following this advice:
>>>>
>>>> https://lists.freebsd.org/pipermail/freebsd-python/2014-October/007606.html
>>>>
>>> Then I am interested in why Marcus is suggesting PYTHON_VERSION rather
>>> than
>>> DEFAULT_VERSIONS. I am currently doing a build run for a separate set
>>> to do
>>> Python 2.7 as the default version for the few ports I need that
>>> require it, but I am using DEFAULT_VERSIONS.
>>
>> I think, you are misinterpreting something or I am not getting what you
>> both are up to.
>>
>> The advice from the mail in October handles the following situation:
>>
>> 1) Joe User uses Python X as default
>> 2) Joe User wants to install a port PP for Python Q
>> 3) To avoid conflicts, Joe User has to
>>   a) install PP with PYTHON_VERSION=Q - or -
>>   b) set up a poudriere with with Python X as default
>> (DEFAULT_VERSIONS=...)
>>      and a make.conf entry setting PYTHON_VERSION=Q
>
> This sounds like my situation where X=2.7 and Q=3.4. Also, I'm using
> poudriere per 3b.
>
>> This is necessary to avoid build and installation conflicts for
>> packages, which support different Python versions (X and Q) at the
>> same time, packages, which  can be installed concurrently, etc.
>
> Yes, this is what I'm after. Am I doing it wrong, or am I tripping over
> real problems/limitations of the ports machinery?

Well, you are tripping over the b2) issue, you already outlined. Its
scope is a bit stretched, though, since I did not consider the
PYTHON_MAJOR_VER stuff, I think.

I'll have a look at that the next days and think of something terribly
complex that apparently will fix your issue, but really will only make
the problem less easy to reveal and occur more randomly based on the
daytime and last command executed :-).

Cheers
Marcus