Fix the bug found by Izidor ('NoneType' object has no attribute 'get_due_date_constraint').
FYI, It seems that this patch has a negative impact on performances (since constraining due dates are computed dynamically for fuzzy/undefined due date). On "normal" dataset, it is about 3% slower (which represents only 30ms on my laptop), but on big datasets (Bryce-ish datasets), it can go up to 10% (=3s on my laptop). I tried to fix this by updating the value only when required and caching it, but I can't succeed with this approach since during boot time it seems that the method "set_parent" is not used to update the task hierarchy. Consequently, I can't catch all task relations updates and the due date constraints are not correctly set up. I have no idea how to catch task relations updates any other way than that, so I don't know how to fix this. Now, since the impact is quite limited, I guess we could still merge the patch.
Fix the bug found by Izidor ('NoneType' object has no attribute 'get_due_ date_constraint ').
FYI, It seems that this patch has a negative impact on performances (since constraining due dates are computed dynamically for fuzzy/undefined due date). On "normal" dataset, it is about 3% slower (which represents only 30ms on my laptop), but on big datasets (Bryce-ish datasets), it can go up to 10% (=3s on my laptop). I tried to fix this by updating the value only when required and caching it, but I can't succeed with this approach since during boot time it seems that the method "set_parent" is not used to update the task hierarchy. Consequently, I can't catch all task relations updates and the due date constraints are not correctly set up. I have no idea how to catch task relations updates any other way than that, so I don't know how to fix this. Now, since the impact is quite limited, I guess we could still merge the patch.