Re: [boost] Re: Detachable objects in BPL

From: Greg Colvin (gcolvin_at_[hidden])
Date: 2001-02-09 14:32:30

From: Phlip <pplumlee_at_[hidden]>
> Proclaimed Greg Colvin from the mountaintops:
> 
> > What is NULL defined as on your system?  ((void*)0) ?
> 
> I see it's time for a version dump:
> 
>   LinuxPPC
>   g++ 2.95.2
> 
> This line returns an error:
> 
>   char * a = (void*)0;

Of course.

> And these lines...
> 
>   #define DEBUG_(x) cout << WRAP(x) << endl
>   #define WRAP(x) #x
>   DEBUG_(NULL)
> 
> ...return "__null".
> 
> This is an amusing chain of events, and I could look up g++'s '__null' for 
> you, but I don't see how it's relevant. (Then again I had no idea one 
> could do what Boost does with traits.)

I strongly suspect that the type of __null is indeed relevant.

> Here's the function to make the goods. I made the names up to protect the 
> guilty:
> 
>       static organ_ptr_t
>       makeOrgan(char const * name)
>       {
>       if (name)
>       {
>       organ_ptr_t result (new organ);
> result -> m_strName = name;
> return result;
> }
> else
> return organ_ptr_t (NULL);
>       }
> 
> The result eventually passes to Python, and that (after the line aOrgan == 
> None) enlightens us with this:
> 
>     if aOrgan.getName() == "":  return
> RuntimeError: Converting from python, pointer or smart pointer to <5organ> 
> is NULL.
> 
> What's happening is this line fails to detect the NULL as a None:
> 
>     if aOrgan == None:  return
> 
> Then the next line bombs.
> 
> Right now I'm using a sentinel value instead of this NULL to mark a 
> non-existent item, but I'd rather be able to use Python's built in None 
> correctly.
> 
> -- 
>   Phlip                          phlip_cpp_at_[hidden]
> ============ http://c2.com/cgi/wiki?PhlIp ============
>   --  http://www.osdi.org/phpwiki/index.php?LinuxChix --
> 
> 
> 
>