(Note: I add space on my link because the forum block my post if contain links) I found this : http:// stackoverflow.com/questions/18680375/display-column-with-no-record-in-openerp-7-kanban-view

In the view, you have this : <kanban default_group_by="stage_id">

You need to specify how sort the group in your class.

def _read_group_state(self, cr, uid, ids, domain, read_group_order=None, access_rights_uid=None, context=None): pass

_group_by_full = { 'stage_id': _read_group_stage_ids}

You can check example with crm_lead: View example: http:// bazaar.launchpad.net/~openerp/openobject-addons/trunk/view/head:/crm/crm_lead_view.xml (search

<record model="ir.ui.view" id="crm_case_kanban_view_leads">

) Sort example: http:// bazaar.launchpad.net/~openerp/openobject-addons/trunk/view/head:/crm/crm_lead.py (search

_group_by_full = { 'stage_id': _read_group_stage_ids }

)

I notice, your stage_id is suppose to be an M2O (or whatever key you want to group by).

Please Mathieu Benoit! can you show us how _read_group_state function should look like?

here is the his mine: 

def _read_group_state(self, cr, uid, ids, domain, read_group_order=None, access_rights_uid=None, context=None):
        print '_read_group_state'
        access_rights_uid = access_rights_uid or uid
        renting_obj = self.pool.get('mymodule.renting')
        
        renting_ids = renting_obj._search(cr, uid, domain, order=read_group_order, access_rights_uid=access_rights_uid, context=context)
        result = renting_obj.name_get(cr, access_rights_uid, renting_ids, context=context)
        
        result.sort(lambda x,y: cmp(renting_ids.index(x[0]), renting_ids.index(y[0])))
        
        states = {}
        for renting in renting_obj.browse(cr, access_rights_uid, renting_ids, context):
            states[renting.id] = renting.state or False
        
        return result, states

that return next error : AssertionError: M2O-like pair expected, got u'draft'