Fatal Python error: Interpreter not initialized (version mismatch?)

[Dan Parisien]

Fredrik Lundh wrote:

> Dan Parisien wrote:
> > If I do not use a thread for the server loop, it does not give the
> > error. The method is called fine, the data decrypted, then unpickled and
> > life is good. There is a problem however... I have no ____ing clue why
> > that happens. Does anyone in python devel know what triggers that error
> > msg?
> 
> it's triggered if Py_InitModule is called (usually by an extension
> module) before the interpreter itself has been initialized.
> 
> on Windows, this can happen if a new version of the interpreter
> picks up an extension linked against an older version of the Python
> DLL, and that DLL is still present.  The runtime linker will then load
> the old DLL (without calling Py_Initialize), and the module ends up
> calling the wrong version of Py_InitModule, in an uninitialized DLL.
> 
> Cheers /F
> 
> 

Thank you for your quick reply.

I grepped the python source code and found that output error. I've tested 
the module under many conditions and it works fine. It only crashes if the 
socket server is in its own thread and each connection has their own thread.

I hope this is a flaw in my code and not in python because I rely on having 
one or more socket servers in their own thread on a listen loop (which I 
had to remove because of this weird problem).

The module in question was compiled on a Mandrake Linux 7.2 system using 
gcc 2.95.3 and swig 1.3u-20001002-1724 (Alpha 3) (ha! say that 3 times 
fast).

Except for this newsgroup, are there other areas where I can get help on a 
subject like this?

Thank you again (this will probably never end ;)
Dan