Re: [Adonthell-devel] Python weirdness (a real weirdy one) - please test

[Alexandre Courbot]

> For some reason, I think that there are some bugs in the python
> callback stuff. Because with your gfxtest code, I get exactly the same
> error that I get with inputtest.py. (Although that could also be
> something OS X related). Anyway, I am using 2.2.2 as well. On running
> gfxtest.py, I get a black window with a little white rectangle. After
> pressing a key, I get:

Don't know if it's the same problem as yours, but I just solved my error. And 
of course it was my fault. :p Looking forward to know whether it makes you 
have better karma with MacOS X.

There was actually two errors of mine. The first one is that the callback in 
gfxtest didn't return an integer (now a boolean) to inform the manager 
whether it caught the event or not. Making the callback return an integer 
solves the problem. And the problem was - precisely - a TypeError involving 
an integer. So obviously it was that error that delayed.

The other error was that I didn't caught this error. In 
python::retrieve_instance<int> I make no check about whether things went 
right or not. It is fixed, and now such errors are displayed in the console 
in time, and doesn't bother SWIG wrappers anymore. Good!

So I've put lots of show_traceback after my Python operations. I don't think 
they will cost much. show_traceback is not more than a call to 
PyErr_Occured() when things went right, and I doubt it does much more than 
just returning a Python variable.

My investigation did more than just fixing this, since I discovered another 
bug in the callback code: the reference to the function was INCREF'ed one 
time too much, causing a slight memory leak.

Alex.
-- 
http://www.gnurou.org