Re: trivial fix in fd.c

"Qingqing Zhou" <zhouqq(at)cs(dot)toronto(dot)edu> writes:
> What's the rationale of designing resowner
> APIs like this:

> ResourceOwnerEnlargeABC();
> ResourceOwnerRememberABC();

> Is that because sometimes we don't allow any elog in
> ResourceOwnerRememberABC()?

Yeah. In a number of cases, ResourceOwnerRememberABC is done after
inserting the resource into some data structure, and an elog there
would leave the data structure in an inconsistent state. Conversely,
we might fail to insert the resource into the other data structure,
so just inserting it into the resowner first isn't better.

regards, tom lane