Re: [HACKERS] No-op case in ExecEvalConvertRowtype

Ashutosh Bapat <ashutosh.bapat@enterprisedb.com> writes:
> While I agree that we can remove indesc->tdtypeid ==
> outdesc->tdtypeid, I am not sure whether it should be replaced by
> !indesc->tdhasoid && !outdesc->tdhasoid.

No, that was overly conservative; the correct test is that the tdhasoid
settings must be equal.  Reading Robert's commit message for 3838074f8
and mine for 3f902354b might clarify this.

> If that's required, it seems
> to be a bug that needs to be fixed in earlier braches as well.

It's not a bug in older branches, because the tdtypeid comparison
was enough to guarantee the same tdhasoid values.
        regards, tom lane