LinuxQuestions.org - PyGTK TreeView segmentation fault on expand

I have recently begun a project using PyGTK, and part of my planned interface has a gtk.TreeView showing a portion of the filesystem. Now, rather than load the entire FS structure into the tree right from the beginning, I made a lazy tree by adding blank children to rows representing directories, and connected a call to fill in the data when a row with blank children was expanded.

Now, this works all fine and well, in general. I can browse my entire filesystem this way. But I noticed that hitting '*' (which I believe is a call to expand_all(), though the documentation does not say this explicitly) on a row representing any empty directory causes a segmentation fault.

This problem can be alleviated by not removing the blank child after attempting to add directory contents, but I would like to avoid this approach.

My suspicion is that expand_all() assumes that there are children present, and when I remove the blank row (after attempting to add any subdirectories and files), it does not check to make sure there are still children.

So I suppose I have a couple questions. First, can anybody confirm my suspicions? Secondly, is this a PyGTK bug, or am I doing something that simply should never be done? Finally, do you see any way to fix this problem?

Code:

def onExpand(self, view, iter, path):

        """Add directory contents on first expansion of its entry"""

        sorted = view.get_model() # TreeModelSort

        iter = sorted.convert_iter_to_child_iter(None,iter)

        store = sorted.get_model() # TreeStore

        child = store.iter_children(iter)

        cpath = store.get(iter,1)[0][1:] # Hidden column with fs path info

        if store.get_value(child, 0) is None:

                sorted.addDir(cpath,iter)

                store.remove(child)

If any other code is necessary, I can provide my entire program. However, I think I've isolated the problem to that last line (execution continues beyond it, so it's not exactly the line itself), and want to hear any recommendations for how to still remove that blank child without causing expand_all() to fall on its face.

Thanks in advance,
Chris