Re: Further news on Clang - spurious warnings

Peter Geoghegan <peter(at)2ndquadrant(dot)com> writes:
> On 5 August 2011 20:07, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> If I were trying to get rid of this warning, I'd be wondering why
>> ReplNodeTag is a distinct enum in the first place.

> Indeed, that doesn't seem to be justified anywhere, and seems to be a
> violation of the abstraction of Node as a pseudo base class which
> would have broken any "downcasting" that we might have attempted to
> do. Since ReplNodeTag wasn't being used directly, just its enumeration
> constants, simply moving the constants to the global, generic NodeTag
> enum fixes the issue.

> That is what the attached patch does.

I did this plus moving replnodes.h into a saner spot: if they're
full-fledged Nodes, they ought to be defined in src/include/nodes/.

I did not renumber the existing node tags as your patch suggests.
We might want to do that at some point to make a bit more breathing
room, but I think it's too late in the 9.1 cycle to be doing that
in 9.1. People have probably already built third-party code that
incorporates values like T_ReturnSetInfo.

regards, tom lane