gtg-contributors team mailing list archive

[Lionel Dricot <ploum@xxxxxxxxx>]

Hello,

I've worked on the dethreadization branch and, so far, it looks
relatively good.

I've worked on the filtered tree to ensure that we have atomic state
change with one state_commit = one signal (it's what gtk is expecting).
__add_node and __delete_node have been rewritten.

There are a few places where I cannot do that and I'm printing a warning
or raising an exception. (they are related to nodes with a parent having
multiple paths).


Error that can arise are fairly easy to spot in the code :

- A commit_state is not immediately followed by a callback
- A callback is not immediately preceded by a commit_state  (well, I
want to refactorize to put the callbacks in the commit_state method)
- After a commit_state, we reuse some values from before the commit
(like the tmp_nodes, the tmp_vr or any local variable)


Doing that work, I broke a bunch of tests in test_liblarch. Still 7
tests are still not passing (so it's low hanging fruit for those wanting
to help). Also, the whole suite is crashing but I think it's not related
to my work.

(Oh, btw, Thread_protection has to be disabled in treemodel.py to be
able to run the tests. I don't know why)

I'm not able to have any issue anymore with -s standard. There are still
some threads race with -s bryce, showing that, somewhere, we might
access the gtk.TreeView with a thread that should not.

The error is :
gtg: Fatal IO error 11 (Ressource temporairement non disponible) on X
server :0.0.


Lionel

PS: as soon as all the tests are passing, I plan to merge that in trunk.
Do you think it's a good idea ?