Re: REINDEX CONCURRENTLY causes ALTER TABLE to fail

On Thu, Jul 25, 2019 at 07:31:13PM +0900, Michael Paquier wrote:
> I have been looking at this code today and extended the test cases
> with more expression and predicate indexes to stress more the patch,
> and it happens that I am still stuck for with makeIndexInfo().  The
> header comments of makefuncs.h and makefuncs.c mention primitive
> nodes, so I would need at least to update that part to mention
> execution nodes.  Still I'd rather not have people scream at me as
> that could be considered invasive in terms of included dependencies
> for makefuncs.c :)

After an extra round of polishing, I am finishing with the attached
and the following change for makefuncs.h:
  * makefuncs.h
- *   prototypes for the creator functions (for primitive nodes)
+ *   prototypes for the creator functions for various nodes

Any objections with this version?  Please note that the predicates and
expressions are fetched directly from the catalogs, and that we does
not rely on the relcache anymore.
--
Michael