> Bruce Momjian <pgman@candle.pha.pa.us> writes:
> >>> I always rip that stuff out. We do an initdb for every release. What
> >>> could break by renumbering those nodes?
> >>
> >> Actually, since we don't store node type numbers on disk (just symbolic
> >> names), there's probably nothing that could break as long as people do a
> >> full recompile when they update. It's no worse than changing the fields
> >> of a widely-used struct, I suppose. I was just being paranoid.
>
> On third thought, there is some value to preserving the node numbers:
> when we see a bug report from the field like
>
> ERROR: ExecEvalExpr: unknown expression type 108
>
> we don't have to resort to excavating back versions of nodes.h to figure
> out which node type is meant.
If we do that, we have to only add entries to the end and not put them
with other appropriate symbols, and we have to keep them around forever.
Doesn't seem worth it to me.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026
