Re: [cgkit-user] segfault help?

[Matthias B. <mat...@gm...>]

On 01.05.12 21:57, Alejandro Aguilera Martínez wrote:
> $ python tst.py 
> Traceback (most recent call last):
>   File "tst.py", line 7, in <module>
>     print (testmod.err(-1))
> RuntimeError: Test error
> 
> I guess this was expected to throw a seg fault and it didn't. Just after
> that I ran "from cgkit.all import *" in a python interpreter and it
> threw a seg fault. What should I do now?

Well, looks like just throwing an exception is not the problem, there
must be some other condition to trigger the segfault and we still need
to find what that is.

We could now reduce the cgkit module step by step until we reach this
very simple example. While doing so, there must be a point where we
switch from a module that does not work (current cgkit) to a module that
does work (the simple example). When we have found that transition point
we might have found the problematic piece of code.

But before doing that, let's change the example a little bit so that it
actually matches what's done in cgkit. I have attached an updated
wrapper.cpp. This version throws a custom exception and registers an
exception translator which is also what's done in cgkit. This means
instead of receiving a RuntimeError in Python, you should instead see a
ValueError.

If that example still works, you could try and remove stuff from
wrappers/py_wrapper.cpp (which is the main wrapper file). The main
function is the one at the bottom which begins with
BOOST_PYTHON_MODULE(_core). Initially, you could comment out everything
in that function except for the call to class_vec3(). That means, the
generated module will only contain the vec3 class. If that results in a
module that does not segfault anymore, you could try and add stuff back
in step by step and see when the segfault comes back.

Cheers,

- Matthias -