browse_null should have __iter__ method to prevent looping problems

[Ronald Portier (Therp)]

When a method that is expecting a browse_record receives a browse_null record, a looping condition might occur, that will lead to a server becoming totally unresponsive, while CPU utilization goes to 100%.

We encountered the problem in a line like the next, where mail_message is supposed to be a browse_record:

if field in mail_message:

because mail_message is in fact a browse_null record, this will result in an endless loop. Well, theoretically not endless, but __getitem__ will be called with increasing index values, in an attempt to find the item with name 'field'.

Look at this discussion:

http://stackoverflow.com/questions/926574/why-does-defining-getitem-on-a-class-make-it-iterable-in-python

The fix is simple. Give browse_null the same __iter__ method as browse_record:

    def __iter__(self):
        raise NotImplementedError("Iteration is not allowed on %s" % self)

It might be argued that modules should not have problems where browse_null objects are treated as browse_record objects. However, it is better to have a clear exception when using a specific function in a module, than a problem hanging a server with no message at all.