Jython, GILs and object locking.

[Andrew Dalke]

Carl Banks:
> Well, like the guy said, that would mean you have to guard the
> reference counter with locks.  That seems far too expensive to me.

The lock is only needed when the reference count is changed,
right?  That is, code with no ref counting locks don't have the
lock overhead.

> I would say the two things are practically irreconcilable.  Either we
> rewrite the C libraries not to depend on order of deallocation (I
> don't think depending on it is the most robust thing to do anyways),
> or keep things the way they are.

I can't rewrite this large third-party C library I use.  (Well, I could,
but it would take me a long time.)  So I guess that means we need
to keep things the way they are?

                    Andrew
                    dalke at dalkescientific.com