[Bf-committers] Blender + python 2.2.2
Meino Christian Cramer
bf-committers@blender.org
Mon, 21 Apr 2003 11:04:12 +0200 (CEST)
From: Michel Selten <michel.s@home.nl>
Subject: Re: [Bf-committers] Blender + python 2.2.2
Date: 19 Apr 2003 16:05:32 +0200
Hi,
thanks for answering my question...
> On Wed, 2003-04-16 at 21:55, Meino Christian Cramer wrote:
> > using Linux:
> > How do I have compile Python, that all symbols needed are found
> > while using the Python scripting interface of a current blender
> > cvs snapshot ?
>
> Erm, you're referring to messages like:
>
> ImportError: 'No module named _Blender'
No, I dont.
As previously reported the problem is as follows:
I am using "povanim" to render objects. PovAnim is a exporter script
to PovRay 3.5.
Currently I use the version Povanim 226 v.11 (the "226" means, it is
for Blender 2.26+.( Attachment to this mail removed, since it is too
long (~400k))
I have an old cvs snapshot, which runs well with this script, so I
think, the script is not the problem.
All cvs-snapshots, which build blender using shared libs has this
bug.
Using '--enable-blenderstatic' does not work, the result is a dyn. built
blender.
A current snapshot produces the following message on the terminal:
guessing 'blender' == '/usr/local/bin/blender'
Created window
Mapped window
Traceback (most recent call last):
File "/home/mccramer/data/pool4/HDA8/.process/lanc_povanim226_11.py", line 39, in ?
import povanim226_11
ImportError: No module named povanim226_11
Traceback (most recent call last):
File "/home/mccramer/data/pool4/HDA8/.process/lanc_povanim226_11.py", line 39, in ?
import povanim226_11
File "povanim226_11.py", line 45, in ?
ImportError: /usr/lib/python2.2/lib-dynload/math.so: undefined symbol: PyExc_OverflowError
Blender quit
(I tried to start the script twice, sometimes the second ALT-P
starts the script with my old cvs-snapshot. It seems, that the first
ALT-P leads to a "harmless" error, which is based on not finding the
*.pyo-module of PovAnim -- which is found in turn with the second
hit of ALT-P, as it did with the old cvs-snapshot.
The second part of the message reveals the "real error": an
undefined symbol.
Python itsself has no problems with the math module.
The error can also be produced with a simple script like:
from math import *
Hitting ALT-P produces:
guessing 'blender' == '/usr/local/bin/blender'
Created window
Mapped window
Traceback (most recent call last):
File "Text", line 1, in ?
ImportError: /usr/lib/python2.2/lib-dynload/math.so: undefined symbol: PyExc_OverflowError
This time another symbol is undefined...
Doing the same inside a python shell works without problems.
I built Blender on a Linux system especially build for this task
alone -- which takes me 6 days: It was compiled from bottom up with
gcc-3.2.2. and is based on instruction which can be found in the
Linux-from-scratch-Manual.
I heard in this list, that gcc-3.2.2 would have "problems" ... but
after building up a whole Linux box including the complete KDE 3.1.1a
and gnome 2.2. suite whithout any problems, I cannot believe this
anymore...
Forthermore I heard, that building Blender with autoconf (which I
use) is buggy. But asking this mailing list for help with the NaN
Makefile system results in no answer at all.
So I see a problem here:
A "normal" person, which only have the knowledge to build up a Linux
box from scratch like me cannot get the Makefile system to run
without any help, which seems not to be available neither from
dokumentations nor from this list.
The Autoconf system is said not to be in a state, which guarantees a
reliable blender build in the sense of reproduceable and therefore
helpful error reports.
I neither want to flood the list with useless error reports like
"something goes wrong here" nor do I want to get answers like "use
the NaN makefiles" without any further help how to get through that
dschungle.
So, what can I do...how can I proceed ?
Kind regards,
Meino
> If that's the case, then don't worry :)
> These happen to everybody who's building blender.
> Here's why:
> intern/python depends on some external variables which are created in
> source/blender/bpython. The bpython library is not built yet, and it's
> in the blender philosophy to not let any source file in intern know
> about a file in source. Even though these messages look scary, the end
> result is that blender should be linked successfully. When blender is
> executed and a script is started, everything works as intended since the
> modules are found at run-time.
>
> With regards,
> Michel
>
>
> _______________________________________________
> Bf-committers mailing list
> Bf-committers@blender.org
> http://www.blender.org/mailman/listinfo/bf-committers
>