On 05/08/2014 12:21 AM, Tom Lane wrote:
> If Craig has a concrete argument why all GUCs should be accessible
> to external modules, then let's see it
Because they already are.
The only difference here is that that access works only on !windows.
I agree (strongly) that we should have a better defined API in terms of
what is "accessible to external modules" and what is not. However, we
don't, as you stressed just that in a prior discussion when I raised the
idea of using -fvisbility=hidden to limit access to some symbols.
Given that we don't have any kind of exernal vs internal API division,
why pretend we do just for one platform?
As for just GUCs: I suggested GUCs because GUCs are what's been coming
up repeatedly as an actual practical issue. I'd be quite happy to
PGDLLEXPORT all extern vars, but I was confident that'd be rejected for
aesthetic reasons, and thought that exporting all GUCs would be a
reasonable compromise.
-- Craig Ringer http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services