[C++-sig] Details of Boost.Python Py_Finalize issue?

[Haoyu Bai]

Hello,

As suggested by Stefan, I'm looking into the Boost.Python Py_Finalize
issue. According to Boost.Python manual, the issue is:

"Currently Boost.Python has several global (or function-static)
objects whose existence keeps reference counts from dropping to zero
until the Boost.Python shared object is unloaded. This can cause a
crash because when the reference counts do go to zero, there's no
interpreter. In order to make it safe to call PyFinalize() we must
register an atexit routine which destroys these objects and releases
all Python reference counts so that Python can clean them up while
there's still an interpreter."

However, until now I can't reproduce the problem. I tried to modify
the exec.cpp in BPL test suite to call Py_Initialize and Py_Finalize
many times in a loop ( http://pastebin.com/f77279354 ), and also tried
examples provided in some report like this:
http://d.hatena.ne.jp/floth/20090525/1243247990 . But neither of them
can cause a crash or observable memory leak on my machine (Arch Linux
with gcc 4.4, Python 2.6 and Boost 1.39).

I also checked BPL source code, but can't spot out any suspicious code
that may cause problems.

So could anyone share your experience with this issue, provide some
test case or anymore information about this issue? I will try to solve
it.

Thank you!

-- 
Haoyu Bai
School of Computing,
National University of Singapore.