[C++-sig] Reference counted object return value policy

[Jean-Sébastien Guay]

Hello again,

> So you're saying that exposing a thin wrapper that would return a 
> ref_ptr<osg::Node> instead of osg::Node would work? 

Continuing my (mostly bind) experimentation, I tried this:

osg::ref_ptr<osg::Node> readNodeFileWrapper(const std::string& filename)
{
     return osg::ref_ptr<osg::Node>(osgDB::readNodeFile(filename));
}

BOOST_PYTHON_MODULE(_osgDB)
{
     def("readNodeFile", readNodeFileWrapper);
}

That gets me a python.exe crash like the other experiments did, running 
this way from the Python interpreter:

import osg
import osgDB
n = osgDB.readNodeFile("cow.osg")

Note that this works:

import osg
n = osg.Node()

so I assume my osg::Node wrapper is at least minimally functional... 
Though I have no way of knowing if n is actually properly constructed or 
if it's enclosed in a ref_ptr... My osg::Node wrapper starts like this, 
for reference:

class_<Node, bases<Object>, ref_ptr<Node> >("Node")
     /* ... */
;

Another weird thing, if I change the above thin wrapper to hard-code the 
filename:

osg::ref_ptr<osg::Node> readNodeFileWrapper()
{
     return osg::ref_ptr<osg::Node>(osgDB::readNodeFile("cow.osg"));
}

I don't get a crash, but OSG reports it can't load a file called "α" 
(lowercase greek alpha), which seems to indicate that somewhere, the 
string got mangled somehow? I don't understand how that can be since 
it's a direct call into OSG from C++ code, nothing else involved.

Has anyone seen something like that before? Thanks in advance,

J-S
-- 
______________________________________________________
Jean-Sebastien Guay    jean-sebastien.guay at cm-labs.com
                                http://www.cm-labs.com/
                         http://whitestar02.webhop.org/