On a separate note, I was having trouble getting the latest versions of cells and celtk to play nicely together. I reverted back to cells from March to work on this stuff since that's when I was last working on my expander widget. I'll investigate the problems and try to provide some more useful input later this week.
This rings a bell. I tripped up Peter with a change that tightened things up. Peter, do you remember what that was? I think it was the change distinguishing the list of "askers" and the idea of the dependent Cell, such that ask cycles got detected even where one had used without-c-dependency (perhaps by calling fm-other).
Yep, I remember there was something like that, but IIRC you modified the cycle-detection code in cells to address this issue.
Off the top of my head (Disclaimer: I might mix things up here), Ken identified some recently introduced constrained in cells which collided with the way menus were instantiated in cells-gtk.
Andy, if you could send us a more detailed description of the problem, that might help trigger our recollection from the cells-gtk3 migration.
Peter