Re: shadow variables - pg15 edition

On Fri, 19 Aug 2022 at 11:21, Justin Pryzby <pryzby(at)telsasoft(dot)com> wrote:
>
> On Thu, Aug 18, 2022 at 07:27:09PM +1200, David Rowley wrote:
> > Any objections to pushing this to master only?
>
> I won't object, but some of your changes are what makes backpatching this less
> reasonable (foreach_current_index and newtype). I had made these v15 patches
> first to simplify backpatching, since having the same code in v15 means that
> there's no backpatch hazard for this new-in-v15 code.

I spent a bit more time on this and I see that make check-world does
fail if I change either of the foreach_current_index() changes to be
incorrect. e.g change the condition from "> 0" to be "== 0", "> 1" or
"> -1".

As for the newtype change, I was inclined to give the variable name
with the most meaning to the one that's in scope for longer.

I'm starting to feel like it would be ok to backpatch these
new-to-pg-15 changes back into PG15. The reason I think this is that
they all seem low enough risk that it's probably more risky to not
backpatch and risk bugs being introduced due to mistakes being made in
conflict resolution when future patches don't apply. It was the
failing tests I mentioned above that swayed me on this.

> I am opened to presenting the patches differently, but we need to come up with
> a better process than one person writing patches and someone else rewriting it.

It wasn't my intention to purposefully rewrite everything. It's just
that in order to get the work into something I was willing to commit,
that's how it ended up. As for why I did that rather than ask you to
was the fact that doing it myself required fewer keystrokes, mental
effort and time than asking you to. It's not my intention to do that
for any personal credit. I'm happy for you to take that. I'd just
rather not be batting such trivial patches over the fence at each
other for days or weeks. The effort-to-reward ratio for that is
probably going to drop below my threshold after a few rounds.

David