Gramps 3.4.9 problem with ubuntu 16.04, which may be related to GTK 2.24.30

[Enno Borgsteede <ennoborg gmail com>]

Hello,

This is a repost from a thread that I started on ubuntu forums at

http://ubuntuforums.org/showthread.php?t=2329261

Forum post below:

For a couple of reasons, I'm still using Gramps 3.4.9, and not the GTK+ 
3 based 4.X series. Gramps 3.4.9 works fine for me on older ubuntu 
versions, but not when I upgrade to 16.04.

In Gramps, I normally use a view where persons are grouped by surname, 
where each surname is a node that can be expanded to see all individual 
persons with that name. Similar views exist for sources, where a node 
can be a book, hiding all entries found in that book.

Anyway, in all ubuntu versions up to 15.10, this tree view works very 
well, but in 16.04 it does not. It actually has two problems:

1. When I expand the last surname in the view, Gramps generates an 
exception. When I add log statements to see what happens, it shows that 
in a tree with two surnames, where only nodes 0 and 1 exist, a call is 
made for node 2, which does not exist. I can demonstrate that in a tree 
with a John A and Jane B.

2. My real tree has more than 8000 different surnames, and when I expand 
a surname near the end, like any surname starting with a Z, Gramps 
freezes for a couple of seconds. This also happens when I simply select 
a person near the end of the alphabet for editing. And in this case logs 
show that, before expanding say node 8000, calls are made for all nodes 
from 0 to 7999 to check whether these nodes still exist in the display tree.

When I start Gramps with the -v option, it lists the versions of some 
loaded Python modules, and of Python itself, and between ubuntu 15.10 
and 16.04, I see that GTK has changed from version 2.24.28 to 2.24.30, 
and Python itself from 2.7.10 to 2.7.11. And being a Gramps developer 
myself, I know that GTK is handling the person view, I suspect that.

And now the big question is: Does anyone know how to test this? I can 
add log statements to Gramps itself, but I prefer to see what GTK+ is 
doing. I like to know why it tries to select the non existing node #2 in 
a 2 surname tree, and why it needs to walk the whole tree to find node 
#8000, where the older version doesn't do either, and only walks the 
whole tree when it is really needed, like when a surname for a person is 
changed, and the whole thing needs to be sorted again.

I know how to compile a GTK version myself, and it's no big issue to add 
log statements either, but I don't want to replace the system GTK+ 
version 2, because many other applications still depend on it, and I 
don't want to corrupt my whole system. So, what I like to know is how I 
can push Python to work with a hacked GTK+ version 2 that sits in my 
home folder.

Ideas, anyone?

thanks,

Enno

P.S. I can not prove that GTK really is the problem. It may also be 
Python making GTK loose focus in tree views somehow.