Re: Be strict when request to flush past end of WAL in WaitXLogInsertionsToFinish

Hi,

On Sun, Oct 25, 2026 at 6:35 PM Paul Kim <mok03127(at)gmail(dot)com> wrote:
>
> On Mon, Oct 19, 2026, Bharath Rupireddy
> <bharath(dot)rupireddyforpostgres(at)gmail(dot)com> wrote:
> > I looked at the TAP test, and I found it a bit hard to follow given
> > the many hard-coded values, [...]
> > Would something like the below work instead [1]?
>
> Yes, your test is easier to follow, so v7 uses it as 0002 in place of
> the test_misc one. 0001 is unchanged from v6.
>
> With 0001, the test passed 5 runs out of 5 on master. Without it, all
> 3 runs failed, with pg_switch_wal() returning the boundary + 40. Since
> data checksums are now on by default, I also checked with pg_waldump
> that no hint-bit FPI lands in the segment the test fills: it holds only
> the 17 logical messages and the switch record.
>
> The test applies to REL_15 through REL_19 as well, with only an offset.

Thanks. I reviewed the v7 patches, and they look good to me. +1 to
back-patch to all the supported branches as of the commit. I will
leave the commit message wording to the committer's taste. I marked
this RfC: https://commitfest.postgresql.org/patch/7252.

--
Bharath Rupireddy
Amazon Web Services: https://aws.amazon.com