kicad-developers team mailing list archive

[Lorenzo Marcantonio <l.marcantonio@xxxxxxxxxxxx>]

On Thu, May 02, 2013 at 03:18:43AM -0500, Dick Hollenbeck wrote:
> This is an application bug, one which failed to properly dovetail object life cycles in
> the dual language environment.  I will get with Miguel on it in the next couple of weeks.

It's inherently difficult. I don't know if it can be done without
heavily massaging pcbnew classes to let them know about the python
interface.
> To cause the bug, my thinking is all you have to do is:
> 
> 1) load the board into pcbnew.
> 
> 2) start the python console, which I will call an "interpreter instance", or interpreter
> for short.
> 
> 3) have interpreter get a board reference.
> 
> 4) Use pcbnew to unload the board, possibly loading another one.
> 
> 5) have the interpreter use the old board reference.

Exactly. I did exactly that.

> BTW, JNI would not have been immune to this scenario either.  Like any tool, it can be
> mis-applied.

My opinion is that is hopeless attaching managed to unmanaged code
without trickery in both of them. Trickery in manager code usually is
easier at least :D

-- 
Lorenzo Marcantonio
Logos Srl