Hi Michael,

On Fri, Nov 26, 2010 at 10:34 AM, Michael Platings <[email protected]> wrote:
> In that case I was thinking of a different potential problem :)
> What was the problem exactly?

Look to be an issue with ordering of initialization of static
variables from one NodeKit to another.  The only way to ensure safe
ordering was to place the static variables within the scope of the
singleton method.

> I want OSG to be as thread safe as possible so
> I'm keen to find a solution that solves both issues.

Ones are very real problem that stops you regardless of threading -
all static apps fail at startup.  The other problem will only occurs
in very specific threading usage models and if your unlucky during
startup of the threads that they exactly overlap.  In theory the later
could be an issue, but as yet I've never seen a bug reported that was
traced back to this.

Solving the the possible initialization threading issue should be a
low priority for us - there are many problems that  effect a far wider
set of users.. like the the glsl_julia.osg regression that points to
wider problems.  In terms of possible solutions for this theoritical
threading bug, wouldn't just invoking the singleton at start up be
suffcient? A global static object that is constructed that then
invokes the singleton method would seems to me sufficient to do this.

Robert.
_______________________________________________
osg-submissions mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org

Reply via email to