Re: v13: CLUSTER segv with wal_level=minimal and parallel index creation

On Mon, Sep 07, 2020 at 07:47:09PM -0700, Noah Misch wrote:
> On Tue, Sep 08, 2020 at 10:43:32AM +0900, Kyotaro Horiguchi wrote:
> > At Tue, 08 Sep 2020 09:13:53 +0900 (JST), Kyotaro Horiguchi <horikyota.ntt@gmail.com> wrote in 
> > > At Mon, 7 Sep 2020 02:32:55 -0700, Noah Misch <noah@leadboat.com> wrote in 
> > > > As a PoC, this looks promising.  Thanks.  Would you add a test case such that
> > > > the following demonstrates the bug in the absence of your PoC?
> > > > 
> > > >   printf '%s\n%s\n%s\n' 'log_statement = all' 'wal_level = minimal' 'max_wal_senders = 0' >/tmp/minimal.conf
> > > >   make check TEMP_CONFIG=/tmp/minimal.conf
> > > 
> > > Mmm. I was close to add some tests to 018_wal_optimize.pl but your
> > > suggestion seems better.  I added several ines to create_index.sql.

After looking closer, I've moved the test to reindex_catalog.sql; see that
file's header comment for the reasons.  One now needs this command:

  printf '%s\n%s\n%s\n' 'log_statement = all' 'wal_level = minimal' 'max_wal_senders = 0' >/tmp/minimal.conf
  make check-tests TESTS=reindex_catalog TEMP_CONFIG=/tmp/minimal.conf

> > > > Please have the test try both a nailed-and-mapped relation and a "nailed, but
> > > > not mapped" relation.  I am fairly confident that your PoC fixes the former
> > > > case, but the latter may need additional code.
> > > 
> > > Mmm. You're right. I choosed pg_amproc_fam_proc_index as
> > > nailed-but-not-mapped index.
> > 
> > I fixed a typo (s/staring/starting/).
> 
> At a glance, this looks reasonable.  If a closer look doesn't reveal problems,
> I'll push this.

RelationBuildDesc() calls RelationInitPhysicalAddr(), so RelationBuildDesc()
can stop calling RelFileNodeSkippingWAL().  The attached version makes it so,
and I plan to push it.