Re: REINDEX INDEX results in a crash for an index of pg_class since 9.6

Hi,

On 2019-04-30 15:11:43 -0400, Tom Lane wrote:
> Andres Freund <andres(at)anarazel(dot)de> writes:
> > On 2019-04-30 14:41:00 -0400, Tom Lane wrote:
> > > On 2019-04-30 12:03:08 -0700, Andres Freund wrote:
> > > > This is a pretty finnicky area of the code, with obviously not enough
> > > > test coverage. I'm inclined to remove them from the back branches, and
> > > > try to get them working in master?
> > >
> > > I think trying to get this "working" is a v13 task now. We've obviously
> > > never tried to stress the case before, so you're neither fixing a
> > > regression nor fixing a new-in-v12 issue.
>
> > Well, the test *do* test that a previously existing all-branches bug
> > doesn't exist, no (albeit one just triggering an assert)? I'm not
> > talking about making this concurrency safe, just about whether it's
> > possible to somehow keep the tests.
>
> Well, I told you what I thought was a safe way to run the tests.

Shrug. I was responding to you talking about "neither fixing a
regression nor fixing a new-in-v12 issue", when I explicitly was talking
about tests for the bug this thread is about. Not sure why "Well, I told
you what I thought was a safe way to run the tests." is a helpful answer
in turn.

I'm not wild to go for a separate TAP test. A separate initdb cycle for
a a tests that takes about 30ms seems a bit over the top. So I'm
inclined to either try running it in a serial step on the buildfarm
(survived a few dozen cycles with -DRELCACHE_FORCE_RELEASE
-DCATCACHE_FORCE_RELEASE, and a few with -DCLOBBER_CACHE_ALWAYS), or
just remove them alltogether. Or remove it alltogether until we fix
this. Since you indicated a preference agains the former, I'll remove
it in a bit until I hear otherwise.

I'll add it to my todo list to try to fix the concurrency issues for 13.

Greetings,

Andres Freund