On Sun, Nov 30, 2008 at 1:09 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Hm, too bad you didn't respond to my message inquiring about this a
> couple days ago.
My apologies -- my mail filters were overeager, and I didn't see it.
> I'll take a look in a bit. However, the switch in the shutdown function
> was simply wrong, because it was *not* "returning to the caller's
> context". I suspect what this really says is dblink is doing something
> it shouldn't.
Well, dblink is simply calling SRF_RETURN_DONE() when the current
context is the multi-call memory context. We could outlaw that
practice, but that risks breaking out-of-tree SRFs that do something
similar. I think better would be to figure out a more appropriate
memory context to switch into before deleting the multi-call context.
Neil
