Asymmetry in globals __getitem__/__setitem__

[Gregory Ewing]

Robert Lehmann wrote:
> I have noticed there is a slight asymmetry in the way the interpreter 
> (v3.3.5, reproduced also in v3.5.x) loads and stores globals.  While 
> loading globals from a custom mapping triggers __getitem__ just fine, 
> writing seems to silently ignore __setitem__.

I didn't think that using a custom mapping object for
globals was officially supported. Has that changed?

-- 
Greg