Re: [PATCH] Fix leaked hash scans in pg_listening_channels()

[Tom Lane]

I wrote:
> I am wondering why you used MAT_SRF_USE_EXPECTED_DESC?

Ah, never mind that: I find that if you don't, you get

regression=# SELECT pg_listening_channels() ;
ERROR:  return type must be a row type
regression=# \errverbose 
ERROR:  XX000: return type must be a row type
LOCATION:  InitMaterializedSRF, funcapi.c:109

I question whether this aspect of InitMaterializedSRF is
well-designed, but I suppose now is no time to rethink that.

                        regards, tom lane