Re: [PoC] Improve dead tuple storage for lazy vacuum

On Wed, Nov 16, 2022 at 2:17 PM John Naylor
<john(dot)naylor(at)enterprisedb(dot)com> wrote:
>
>
>
> On Wed, Nov 16, 2022 at 11:46 AM John Naylor <john(dot)naylor(at)enterprisedb(dot)com> wrote:
> >
> >
> > On Tue, Nov 15, 2022 at 11:59 AM Masahiko Sawada <sawada(dot)mshk(at)gmail(dot)com> wrote:
> > > Thanks! Please let me know if there is something I can help with.
> >
> > I didn't get very far because the tests fail on 0004 in rt_verify_node:
> >
> > TRAP: failed Assert("n4->chunks[i - 1] < n4->chunks[i]"), File: "../src/backend/lib/radixtree.c", Line: 2186, PID: 18242
>
> Actually I do want to offer some general advice. Upthread I recommended a purely refactoring patch that added the node-pointer struct but did nothing else, so that the DSA changes would be smaller. 0004 attempted pointer tagging in the same commit, which makes it no longer a purely refactoring patch, so that 1) makes it harder to tell what part caused the bug and 2) obscures what is necessary for DSA pointers and what was additionally necessary for pointer tagging. Shared memory support is a prerequisite for a shippable feature, but pointer tagging is (hopefully) a performance optimization. Let's keep them separate.

Totally agreed. I'll separate them in the next version patch. Thank
you for your advice.

Regards,

--
Masahiko Sawada
Amazon Web Services: https://aws.amazon.com