Tom Lane wrote:
> A possible objection to this is that most C compilers wouldn't complain
> if a call site is still trying to use the old convention of passing a
> CmdType value. In the core code, there are only four call sites and
> three are in rewriteHandler.c itself, so this isn't much of a problem
> --- but if there's any third-party code such as FDWs that's trying to
> make use of this function for querytree manipulation, there'd be a risk
> of failing to notice the need to update the call.
Failing to notice such changes is easy if the compiler doesn't even
issue a warning, so *some* way to have old code fail (even better if
it's a hard error and not just a warning) would be nice. I'm not sure I
have useful suggestions on how to do it, though, just a +1 to doing it.
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services