Andres Freund <and...@anarazel.de> writes:
> ... in the case the attribute isn't already deformed, the
> following hunk exists:
> /*
> * If the attribute's column has been dropped, we force a NULL result.
> * This case should not happen in normal use, but it could happen if we
> * are executing a plan cached before the column was dropped.
> */
> if (TupleDescAttr(tupleDesc, attnum - 1)->attisdropped)
> {
> *isnull = true;
> return (Datum) 0;
> }
> Which strikes me as quite odd. If somebody previously accessed a *later*
> column (be it via slot_getattr, or slot_getsomeattrs), the whole
> attisdropped check is neutralized.
Good point. Let's remove it and see what happens.
> Tom, you added that code way back when, in a9b05bdc8330. And as far as I
> can tell that issue existed back then too.
I was just transposing code that had existed before that in ExecEvalVar.
Evidently I didn't think hard about whether the protection was
bulletproof. But since it isn't, maybe we don't need it at all.
I think our checks for obsoleted plans are a lot more bulletproof
than they were back then, so it's entirely likely the issue is moot.
regards, tom lane
