Re: corruption of WAL page header is never reported

At Thu, 2 Sep 2021 21:52:00 +0900, Fujii Masao <masao(dot)fujii(at)oss(dot)nttdata(dot)com> wrote in
>
>
> On 2021/09/02 16:26, Kyotaro Horiguchi wrote:
> > I believe errmsg_buf is an interface to emit error messages dedicated
> > to xlogreader that doesn't have an access to elog facility, and
> > xlogreader doesn't (or ought not to or expect to) suppose
> > non-xlogreader callback functions set the variable. In that sense I
> > don't think theoriginally proposed patch is proper for the reason that
> > the non-xlogreader callback function may set errmsg_buf. This is what
> > I meant by the word "modularity".
> > For that reason I avoided in my second proposal to call
> > XLogReaderValidatePageHeader() at all while not in standby mode,
> > because calling the validator function while in non-standby mode
> > results in the non-xlogreader function return errmsg_buf. Of course
> > we can instead always consume errmsg_buf in the function but I don't
> > like to shadow the caller's task.
>
> Understood. Thanks for clarifying this!
>
> > Does that makes sense?
>
> Yes, I'm fine with your latest patch.

Thanks. Maybe some additional comment is needed.

regards.

--
Kyotaro Horiguchi
NTT Open Source Software Center