Workaround focus issue by using the size-allocate signal (Closes: #988951).
Compared to the original focus_path() callback, this updated version no longer disconnects itself, and needs to retain the specified GtkTreePath.
XXX: Memory management, the GtkTreePath is no longer free'd.
I'm not familiar enough with best practices when it comes to callbacks, data they need, etc. or with reference handling…
I kept the patch minimal to show what seems to work fine even if the path
is leaked. Maybe someone has a magic idea to fix that, my fallback plan would be to keep trace of that GtkTreePath
via an extra member of fe->data
(aka. fe_data
), so that the existing one can be freed right before creating a new one… Another kludge, but oh well.
/cc @smcv