Re: [bugle-devel] Windows NULL ptr issue resulting from bugle_thread_lock_lock call

[Bruce M. <bm...@gm...>]

Hi Bryce

This sounds like a new one to me - please do file it in trac.
Unfortunately I'm in the process of moving jobs (and countries), so I
have almost zero time to maintain bugle at the moment.

Based on the stack trace, here's my best guess as to what is
happening. The call to bugle_object_get_current_data is to extract
data about objects stored in the namespace. To determine that, it
first needs to find the current namespace (which is the innermost call
in your trace). That information is stored in the current context
(although a context never changes namespaces, it seems I found it
easier just to have an idea of a current namespace for each context).
The variable ans should point to the piece of the context storage that
holds a pointer to the namespace for that context.

Without being able to attach a debugger myself I think that's about
all I can figure out, but I can guess that the most likely cause will
be that bugle is getting confused about what the current context is -
either because it's not intercepting the WGL calls (entirely possible
- Windows support is nowhere near as tested as GLX), or because the
app isn't making a context current (although I'd expect the app not to
work in that case).

> The data stored in ans comes from obj->views[view] in the
> bugle_object_get_data function.
> The views array contains data initialised by bugle_object_new via the line
> (*info->constructor)(key, obj->views[j]);
> The constructor function pointer is pointing to the globjects_data_init
> function, which initialises globjects_data structs in some memory location
> given a pointer. This pointer is from obj->views[j] in the case above.
> This is where I fear I've made an error, as if the above is correct it would
> mean that ans = bugle_object_get_current_data(klass->parent,
> klass->parent_view); is a void* pointing to a globjects_data. As such the
> *(object **)ans cast doesn't make sense (as far as I can tell).

What I suspect you're missing is that when bugle_object_get_current
calls bugle_object_get_current_data, it passes klass->parent, not
klass. The view it gets back is initialized in bugle_object_class_new.

Regards
Bruce
-- 
Dr Bruce Merry
bmerry <@> gmail <.> com
http://www.brucemerry.org.za/
http://blog.brucemerry.org.za/