oship-dev team mailing list archive

[Roger Erens <roger.erens@xxxxxxxxx>]

On Fri, Apr 30, 2010 at 03:11, Tim Cook <timothywayne.cook@xxxxxxxxx> wrote:
> On Thu, 2010-04-29 at 18:05 -0300, Diego Manhães Pinheiro wrote:
>>
>> Tim I finished the egg build for pyclips. Just try to execute the
>> buildout process again, to see if it works. I did a quick checkout on
>> others OSHIP dependencies. I guess this is the unique that I forgot
>> for the buildout works well with amd64 machines. If you have any other
>> trouble with the buildout process, just ask me. I will glad to help.
> The egg seems to work fine.  However, line 22 in dsedemo.py needs to be
> changed to:
> from clips import clips
>
>
>>
>> I thought on two situations that could solve this problem:
>> 1. Provide a dummy psyco package for amd64 machines, only with stub
>> calls. For example the psyco.full() call will do nothing more than a
>> pass python statement.
>> 2. Provide a buildout recipe to verify which architecture is and then
>> install or not the psyco module.
>>
>>
>> I'm tendentious to choose the last option, but i'm certainly sure if
>> choose the first I could solve more faster.
>> For a while, please consider remove the psyco reference from buildout
>> configuration as workaround to solve this. I'll take care of it in a
>> few days after the continuous integration system was up and running.
>
> I believe that the most expedient way to solve this is to comment out
> psyco.
> Add a comment above it that it should be uncommented for 32 bit
> machines.  Then in the install guide mention it as well.  Just prior to
> running buildout.

That will start cluttering up the installation guide again. I know
it's convenient now, but my feeling is that it will become
inconvenient later on. Diego did a nice job with the egenix test for
UCS2/4, so I think he could do it again with an architecture test.

Cheers