On Wed, Apr 27, 2022 at 1:47 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Thomas Munro <thomas.munro@gmail.com> writes:
> > Hrmph... Are you sure you rebuilt the contrib module? Assuming so,
> > maybe it's failing in a different way for you and me. For me, it
> > always fails after this break is reached in xlogutil.c:
>
> > /* If asked, let's not wait for future WAL. */
> > if (!wait_for_wal)
> > break;
>
> Hmm. For me, that statement is not reached at all in successful
> (make installcheck) runs. In a failing run, it's reached with
> wait_for_wal = false, after which we get the "could not read WAL"
> failure. Usually that happens twice, as per attached.
Ok, that's the same for me. Next question: why does the patch I
posted not help? For me, the error "could not read WAL at %X/%X",
seen on the BF log, is raised by ReadNextXLogRecord() in
pg_walinspect.c. The patch removes that ereport() entirely (and
handles NULL in a couple of places).