Comment 2 for bug 1966905

Revision history for this message
Sebastien Bacher (seb128) wrote : Re: Valgrind memory errors in gnome-shell 42 from accountsservice

Upstream gave some debugging hints

'So the only unref call has a matching ref call pair. This means it shouldn't lead to the object getting freed. But between those two calls the property notify handlers get run, etc. It's at least conceivable to me that one of those handlers is accidentally unrefing the object when it isn't supposed to.

If that is what's going on, how are we going to find out where the handler is? One idea,

You could try putting

g_object_unref (user);

and the top of on_new_user_loaded. This is wrong and adding a bug, but it may also help valgrind to spot the aforementioned, theorized preexisting extraneously unref too. This is because it will make the user object get freed earlier, and valgrind should show where.'

Could you give it a try?