pthreads in C++ with embedded Python

[Jason Tackaberry]

On 11-06-08 06:28 PM, Tom Brown wrote:
> I found that PyEval_ReleaseLock() was necessary to keep the program 
> from hanging. The lock() and unlock() methods were used in a previous 
> attempt to lock/unlock the GIL.

I just tried your example code and indeed it segfaults as is, but works 
fine for me when I comment out PyEval_ReleaseLock().  Perhaps this was 
just a red herring, not actually solving your deadlock problem 
properly?  (That is, replacing one problem with another.)


> I kept banging at this and finally gave up on the GIL. I used a mutex 
> instead. This allowed me to get rid of the Python API calls that dealt 
> with the GIL. It works great in the test program. I'll find out how 
> well it performs in the real program.

I suppose this could theoretically work in your specific example, but if 
the Python module you're calling out to spawns threads you're screwed. :)

Cheers,
Jason.