Re: Remove no-op pull_var_clause flag

Richard Guo <guofenglinux(at)gmail(dot)com> writes:
> On Thu, Jan 22, 2026 at 3:27 AM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> Alexander Pyhalov <a(dot)pyhalov(at)postgrespro(dot)ru> writes:
>>> I was a bit surprised, the patch hasn't landed to master. But flag
>>> somehow slipped in (but only as no-op). The attached patch removes
>>> useless flag.

>> Right, done.

> Ugh... I wonder how this happened, and whether this is the only
> instance of private code sneaking into the PostgreSQL codebase. I'm
> also kind of concerned about the legal risk if this comes from a
> project with a strict license.

Unless there's more here than the one single symbol name, I'm not
worried about legal risk --- it's hard to claim copyright or patent
interests in that much. In any case, it's hard to see postgrespro.ru
suing the rest of us over their own mistake.

> Should we also remove this code from v18?

I thought about it but desisted. There's some epsilon-level risk
that somebody already copied the pull_var_clause call with
PVC_INCLUDE_PLACEHOLDERS into their extension. If so, their code
isn't broken today but would be if we back-patched. Tiny as that
risk is, the benefit of removing the symbol from v18 isn't larger.

Also, while I believe our newly-minted libabigail ABI-checking
infrastructure isn't smart enough to complain about removal of a
macro symbol, it's possible that downstream packaging systems
would notice that and flag it as an inappropriate API change.
Again, the bureaucracy involved in dealing with such a complaint
seems to outweigh the benefit.

regards, tom lane