Re: some more error location support

>> Even if there is some under-the-hood garbage collection, I'd suggest to
>> add a free after the call to ComputePartitionAttrs.
>
> Hmm, I didn't know about free_parsestate().  It doesn't seem to be used
> consistently.  I suppose you'll want to use it when you have a target
> relation that will be closed by it, but otherwise, for DDL commands,
> it's not all that useful.

Probably.

The majority rule (34 make & 22 free) suggest that it is more often use 
than not. I'd suggest to stick to that for consistency & homogeneity.

-- 
Fabien.