Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> writes:
>> I was rather bemused to discover just now that the node-freeing
>> functions in nodes/freefuncs.c are never called from anywhere;
>> in fact, the module hasn't got a single exported entry point!
> Are you sure about this? I thought these things were called from
nm --defined --extern shows that freefuncs.o exports no symbols.
Therefore it's impossible for any outside code to refer to it.
(Hmm, I wonder if any other modules are equally dead?)
> I would perhaps move it to a _deadcode directory and see
> what happens. Or should we enable it?
I certainly don't want to bit-bucket it --- I was just very surprised
that it's not currently being used.
After further consideration I realized that if RelationClearRelation
uses freeObject() to flush rules, then a DROP on a view will exercise
the code. So it's not quite as hard to test as I was thinking before.
I'm leaning towards the "go for it" answer at the moment...
regards, tom lane
In response to
pgsql-hackers by date
|Next:||From: Tom Lane||Date: 2000-01-30 04:18:38|
|Subject: Re: ImmediateSharedRelationCacheInvalidate considered harmful |
|Previous:||From: Bruce Momjian||Date: 2000-01-30 03:18:09|
|Subject: Re: [HACKERS] freefuncs.c is never called from anywhere!?|